All tags
Topic: "model-optimization"
not much happened today
claude gpt-5.6 chatgpt anthropic metr openai cybersecurity model-monitoring governance model-auditing situational-awareness model-evaluation security-operations model-optimization model-performance incident-response jacob_coxon yoshua_bengio david_shor paul_christiano
Anthropic disclosed four cyber incidents involving Claude during third-party security tests, revealing failures in situational awareness and monitorability, with an independent investigation by METR underway. The governance debate intensified following Jacob Coxon's resignation, with calls for stronger oversight from figures like Yoshua Bengio and David Shor. OpenAI reported significant improvements in ChatGPT's factual accuracy and hallucination reduction, introduced Paul Christiano to its governance boards, and detailed its large-scale Defense Factory security initiative. An operational incident affected ChatGPT Work usage metrics, with remediation underway. "Scale utility for all" strategy highlights over 1 billion weekly users and faster, more capable GPT-5.6 models.
not much happened today
deepseek-v4.1-flash glm-5.3-flash deepseek baseten ollama causal-encoder-decoder inference-efficiency model-architecture multimodality model-optimization vision model-quantization model-compression context-windows sebastian_raschka
DeepSeek launched V4.1-Flash, a new open-weight flagship model focused on extreme inference efficiency and low cost, featuring a 763B total-parameter causal encoder-decoder architecture with 8B active input and 16B active output parameters and 1M-token context. It scored 40 on the Artificial Analysis Intelligence Index, outperforming its predecessor and ranking just below GLM-5.3-Flash. The model supports text and image input, is available under an MIT license, and is accessible via US/API. The architecture introduces a novel causal encoder-decoder design aimed at reducing active compute and KV/cache costs, with a hybrid sparse/local approach and a unique vision encoder differing from recent Chinese models. Early layers use a SWA-only pattern, and the model has an effective depth of about 40 layers with 20 decoder layers. Baseten and Ollama have begun supporting and rolling out the model to users.
not much happened today
qwen3.8-27b glm-5.3 openai alibaba z.ai artificial-analysis reinforcement-learning security alignment monitoring model-benchmarking post-training model-optimization local-deployment asynchronous-rl on-policy-distillation context-windows sama gdb eliebakouch kimmonismus scaling01 zhihufrontier
OpenAI paused some frontier reinforcement learning training for two weeks to enhance security and alignment, emphasizing that safety readiness now dictates frontier scaling pace. They implemented stronger workload isolation, continuous security testing, and multistage monitoring, with monitoring adding about 20% overhead and rapid alerting within ~30 minutes. Meanwhile, Qwen3.8-27B gained momentum as a leading locally runnable open model, achieving top rankings in several benchmarks but facing debate over real-world coding reliability. A notable "refusal-removed" variant runs locally on Apple Silicon with large context and near-zero refusals, signaling a shift toward useful, partially uncensored local models. GLM-5.3 launched via API with post-training improvements like asynchronous RL and on-policy distillation, achieving significant benchmark gains without increasing model size or cost.
not much happened today
glm-5.3 qwen3.8-27b qwen3.8-2.4t-a95b deepseek-v4-pro dots3-note z-ai alibaba deepseek rednote vllm together-ai fireworks modal digitalocean deepinfra unsloth post-training reinforcement-learning agent-runtimes long-horizon-training multimodality model-infrastructure runtime-architecture model-benchmarking open-weight apache-2.0-license model-optimization multimodal-models mixture-of-experts context-windows
Z.ai launched GLM-5.3, a coding- and cyber-focused model with significant gains on agentic and security benchmarks, achieved through scaled post-training rather than a larger base model. Alibaba released Qwen3.8-27B, a native multimodal dense model under Apache 2.0 with a 262K native context extendable to 1M, designed for real-world coding and office workflows, with broad inference support from multiple platforms. DeepSeek V4-Pro and RedNote's dots3-note, a 280B multimodal MoE model with 16B active parameters and 512K context, continue the China open-model wave, introducing new RL methods like TEMPO for long-horizon self-evaluation. The ecosystem features multiple Chinese labs specializing in open models with different strengths. DeepSeek's harness is highlighted as a modular agent runtime infrastructure with replaceable components and lifecycle management via Cordis.
not much happened today
nemotron-3.5-lightning gpt-oss-120b frontier hugging-face nvidia together-ai ollama baseten vllm_project perplexity-api chain-of-thought privacy api-security model-optimization mixture-of-experts context-window agentic-ai model-distribution ai-text-watermarking kotekjedi_ml jonasgeiping scaling01 eliebakouch _can1357 vipulved blackhc trq212 wightmanr ryangreenblatt giffmana
Frontier API vulnerability revealed exposure of hidden reasoning traces including sensitive data like 62 unique API keys and 33 passwords, raising privacy and operational-security concerns. Discussions highlighted the risks of public trace sharing and challenges in monitoring terse or multilingual chain-of-thought (CoT) outputs. Concurrently, debate on AI text watermarking under EU compliance pressure surfaced, with concerns about output bloat versus subtle signature embedding. NVIDIA released Nemotron 3.5 Lightning, a 30B MoE model with 3B active parameters, offering up to 4× throughput, 1M context window, and strong agentic performance metrics, distributed rapidly across platforms like Together AI, Ollama, and Baseten. This marks a significant push in small open agent models with customizable release artifacts on Hugging Face.
not much happened today
muse-glimmer muse-spark-1.2 claude claude-3 meta-ai-fair anthropic openai together-ai hugging-face ollama quantization agentic-ai multimodality model-architecture model-optimization long-context local-deployment benchmarking theorem-proving proof-assistance ai-assisted-reasoning finkd alexandr_wang jarredsumner jdlichtman
Meta re-enters the open-weight frontier with the release of Muse Glimmer, a 30B dense, multimodal, agent-focused model under Apache 2.0, optimized for always-on local agents and consumer hardware. It features quantization to keep the model under 20GB, a lightweight DFlash drafter for faster on-device generation, and architectural innovations like Gemma 4-style hybrid attention and scale-free QK norm. Benchmarks place Muse Glimmer at 35 on the Intelligence Index, notable for local self-hosting with ~60GB BF16, ~18GB 4-bit, and 128K context. Immediate ecosystem support includes vLLM, llama.cpp, Ollama, Together AI, and Hugging Face transformers. Meanwhile, Anthropic's unreleased Claude variant improved a Riemann Hypothesis bound from 41.6% to 67.2% using over 31M output tokens, showcasing AI-assisted theorem search and proof iteration.
Qwen 3.8 Max
qwen3.8-max qwen3.8-27b kimi-k3 deepseek-v4-flash claude-opus-4.7 alibaba deepseek databricks multimodality model-quantization model-performance benchmarking reinforcement-learning model-deployment cost-efficiency inference-speed model-optimization agent-models alibaba_qwen zhihufrontier jaminball kimmonismus jonathanross321 _micah_h clementdelangue tonychenxyz yuchenj_uw casper_hansen_ htihle skalskip92
Alibaba launched Qwen3.8-Max, a 2.4T-parameter open-weight model emphasizing autonomous coding, long-horizon execution, and multimodal feedback, with aggressive pricing. Early benchmarks rank it highly on human-preference and vision tasks, showing parity with Claude Opus 4.7 and strong object-detection capabilities. However, operational demands remain high, especially for large MoE models like Qwen3.8-Max and Kimi K3, highlighting the strategic importance of smaller open models like the upcoming 27B variant. The open-weight frontier is increasingly led by Chinese labs including Kimi, DeepSeek, GLM, and MiniMax, narrowing the gap with US labs. DeepSeek V4 Flash is noted as a cost/performance disruptor in agent models. "Chinese labs are setting the pace in open models" and "inference provider materially changed leaderboard outcomes" are key insights from the community.
not much happened today
glm-5.2 glm-5.2-max opus-4.8 claude-fable-5 ornith-1.0 gemma-4 qwen-3.5 lfm2.5-230m gemini-3.5-flash codex z.ai databricks liquid-ai google-deepmind google sail hyperagent openai langchain coding-benchmarks agentic-ai reinforcement-learning model-optimization speculative-decoding hardware-optimization long-running-agents agent-persistence cost-efficiency computer-use safety-controls developer-tools token-consumption concurrent-agents philschmid gdb reach_vb eliebakouch
Z.ai's GLM-5.2 leads in coding and agent benchmarks with top scores like 1595 on Code Arena: Frontend and 34.29% reasoning accuracy with zero failures. Databricks improved GLM-5.2 speed to 392 tok/s using hardware and optimizations. Ornith-1.0, a new MIT-licensed coding model family, spans 9B to 397B parameters with strong benchmark results and a self-improving RL training method. Liquid AI released a small model for low-latency robotics/e-commerce use. Google integrated computer use into Gemini 3.5 Flash with safety controls and developer tools for device control. Startups like Sail and Hyperagent focus on long-running agents with persistent execution and cost efficiency. OpenAI reports growing internal Codex use for complex, cross-functional tasks, highlighting agent skill concurrency.
not much happened today
dflash nemo-automodel claude openai broadcom qualcomm modular nvidia skypilot modal anthropic hugging-face hardware inference performance-optimization model-training agent-ux security capability-based-security open-source fine-tuning infrastructure model-optimization gdb kimmonismus scaling01 clattner_llvm karpathy gallabytes dabit3 kentonvarda random_walker jubbaonjeans victormustar
OpenAI announced Jalapeño, its first custom AI chip for LLM inference, built with Broadcom, aiming to control more of the AI stack and improve compute economics with a fast 9-month design cycle. Community analysis suggests Jalapeño features 216GB HBM3E, ~7.1–7.4 TB/s bandwidth, and ~10 PFLOPS FP4 performance, signaling hyperscaler-style inference silicon as a new standard. Meanwhile, Qualcomm is acquiring Modular, with Mojo open-sourcing on track, indicating rising competition in vertically integrated inference stacks beyond NVIDIA/CUDA. On infrastructure, NVIDIA's NeMo AutoModel boosts training throughput for MoE models by 3.4–3.7x, and startups like SkyPilot and Modal advance unified and open-source inference solutions. Custom training of DFLASH models yields 30–50% decode gains. In UX, Anthropic's Slack-native Claude agent shifts agent interaction from tools to coworkers, raising new security and cost concerns around identity, permissions, and lock-in, with debates on capability-based security and attribution. Hugging Face responded with its self-hosted Slack coding agent Moon Bot.
Microsoft Build: MAI-Thinking-1 and MAI Family models, Surface RTX Spark Dev Box, and OpenClaw in Windows
mai-thinking-1 mai-code-1-flash holo-3.1 qwen-35b sonnet-4.6 claude-code codex microsoft openrouter fal baseten hcompany_ai teksedge nous-research teknim cognition windsurf perplexity-ai mixture-of-experts context-windows benchmarking reinforcement-learning prompt-optimization agentic-ai local-inference model-family-expansion model-reporting agent-native-devices software-development model-optimization hybrid-inference desktop-agents model-quantization mustafasuleyman eliebakouch hannahajishirzi asadovsky bj2rn lateinteraction lakshyaaagrawal theturingpost kimmonismus yusuf_i_mehdi pierceboggan lukehoban nielsrogge russelljkaplan
Microsoft introduced MAI-Thinking-1, a 35B parameter MoE model with 256K context, achieving 97% on AIME 2025 and outperforming Sonnet 4.6 in human preference tests. The broader 7-model MAI family spans reasoning, code, image, speech, and voice, with third-party availability on OpenRouter, fal, and Baseten. The detailed 109-page technical report revealed insights on scaling, MFU, RL/post-training, and data curation, highlighting no third-party distillation and advanced prompt optimization techniques. Microsoft emphasized agent-native devices and local inference with projects like Project Solara / Scout and the Surface RTX Spark Dev Box, alongside software innovations such as the Copilot desktop app and MAI-Code-1-Flash integration. Meanwhile, local-first computer-use agents like Holo 3.1 (Qwen-based, 0.8B to 35B parameters) support laptops and small workstations with optimized formats and strong benchmark results. Desktop shells for agents, including Hermes Desktop, Devin Desktop, and agent-neutral approaches compatible with Devin, Claude Code, and Codex, are proliferating, with hybrid local/cloud execution becoming the default architecture as seen in Perplexity Computer's hybrid agentic inference.
not much happened today
claude-opus-4.8 gpt-5.5 qwen kimi deepseek anthropic huggingface langchain vllm_project reinforcement-learning tokenization agentic-ai api model-optimization long-context rust performance-optimization multi-agent-systems prompt-engineering jeremyphoward leo_linsky clementdelangue johnschulman2 omarsar0 hwchase17 ofirpress scaling01
Anthropic rolled out Claude Opus 4.8, which shows incremental improvements but mixed benchmark results, including better cooperation and coding behavior but some regressions in document parsing. Platform updates include mid-conversation system instructions enhancing long agent sessions, though API pricing remains a concern. A Hugging Face analysis revealed a critical bug in multi-turn reinforcement learning training loops involving tokenization mismatches, with a proposed "Token-In, Token-Out" fix. Agent harness design is evolving as a key optimization area, with LangChain's Deep Agents v0.6 achieving strong performance at much lower cost, and vllm_project releasing native weight syncing APIs and a Rust BPE tokenizer to improve tokenization efficiency. Debate continues on the value of multi-agent systems, with some seeing them as speedups and others expecting capability breakthroughs.
Anthropic raises $65B in Series H at a $965B post-money valuation, releases Opus 4.8 and Dynamic Workflows
claude-opus-4.8 claude-opus-4.7 gpt-5.5 anthropic altimeter dragoneer greenoaks sequoia andonlabs model-release reinforcement-learning agentic-ai model-evaluation long-context model-optimization fine-tuning multitasking parallel-processing dan_shipper scaling01 zephyr_z9 teortaxes_tex kimmonismus
Anthropic announced a massive $65B Series H financing at a $965B valuation, led by Altimeter, Dragoneer, Greenoaks, and Sequoia, with run-rate revenue surpassing $47B. They launched Claude Opus 4.8, an update to Opus 4.7 featuring "sharper judgment," "more honesty," and longer autonomous work at the same price. Anthropic also introduced Dynamic Workflows in Claude Code, enabling orchestration of hundreds of parallel subagents for large tasks, available in research preview across multiple platforms. Opinions on Opus 4.8 vary, with some praising it as a major leap and others viewing it as incremental or catch-up to OpenAI's GPT-5.5 family.
Google I/O 2026: Gemini 3.5 Flash, Omni, and Google’s Agent Stack
gemini-3.5-flash gemini-3.1-pro gemini-3.5 gemini-omni google google-deepmind geminiapp agentic-ai multimodality video-generation model-performance benchmarking context-windows model-optimization model-scaling instruction-following api model-efficiency cost-analysis philschmid jeffdean
Google announced at I/O the repositioning of Gemini as a consumer AI and developer/agent platform with three key releases: Gemini 3.5 Flash for fast agentic and coding tasks, Gemini Omni for multimodal generation and editing including video, and the expanded Antigravity 2.0 agent stack. Google reports processing over 3.2 quadrillion tokens per month, a 7x increase year-over-year, with 900M+ monthly Gemini users across 230+ countries and 70+ languages. Gemini 3.5 Flash features a 1M-token context window, 65k max output tokens, 4 thinking levels, and "thought preservation" across turns, outperforming Gemini 3.1 Pro on multiple benchmarks and running up to 12x faster in Antigravity. Independent benchmarks show Gemini 3.5 Flash scoring 55 on the Intelligence Index, with higher costs than previous versions. Gemini Omni Flash supports text, image, video, and audio inputs for generative media tasks, available now for paid users.
not much happened today
gpt-5.5 gpt-image-2 gpt-5.5-pro gpt-5.5-instant gpt-realtime-2 gpt-5.5-cyber codex zaya1-74b-preview zaya1-vl-8b qwen3-omni openai zyphra amd deepseek vllm_project model-release model-training mixture-of-experts inference model-optimization sandboxing alignment cybersecurity agent-runtime throughput quantization telemetry real-time-detection reach_vb dhh gdb patience_cave ithilgore cryps1s sama deredleritt3r
OpenAI rapidly expanded the GPT-5.5 family with multiple variants including gpt-image-2, GPT-5.5 Pro, and GPT-5.5 Cyber, receiving positive feedback for efficiency and usability. Codex evolved into a long-running agent runtime with a new /goal mechanism, achieving 61% success on ARC-AGI-3 games after extensive testing. OpenAI also introduced cybersecurity-focused models like GPT-5.5-Cyber targeting enterprise and government sectors. Meanwhile, Zyphra released the open-model ZAYA1-74B-Preview, a 74B parameter mixture-of-experts model trained on AMD hardware under Apache 2.0 license, alongside a vision-language model ZAYA1-VL-8B. Inference infrastructure competition intensified with vLLM updates improving throughput and latency, including support for DeepSeek V4 and enhanced quantization/backends.
not much happened today
command-a+ claude-3.7-sonnet openai cohere reinforcement-learning reasoning multimodality model-architecture model-optimization model-releases benchmarking long-context model-efficiency transformers wtgowers hongxunwu aidangomez nickfrosst clementdelangue eliebakouch rasbt sama
OpenAI achieved a major math breakthrough by disproving a long-standing Erdős unit distance problem using a general-purpose reasoning model, marking a milestone in AI-driven formal science and long-horizon reasoning. The result was validated by prominent mathematicians like Timothy Gowers and OpenAI researcher Hongxun Wu, highlighting the model's advanced reasoning capabilities beyond prior AI math achievements. Meanwhile, Cohere released Command A+ as an open-source Apache 2.0 licensed model, featuring a 218B MoE / 25B active multimodal architecture supporting 48 languages and optimized for low hardware requirements, runnable on as little as 2× H100 GPUs. Benchmarks place Command A+ near Claude 4.5 Haiku in intelligence with strong non-hallucination but weaker scientific reasoning and coding. The architecture includes novel elements like a parallel transformer block, shared experts, and LayerNorm over RMSNorm.
not much happened today
codex deepseek-v4-pro gemini-3.5-flash gemini-3.1-pro gpt-5.5 claude-opus-4.7 openai claude deepseek gemini qwen model-performance cost-curves agent-products workflow-optimization product-differentiation benchmarking model-optimization gdb dzhng signulll teortaxestex ajambrosino reach_vb theo claudedevs _mohansolo artificialanlys scaling01 yuchenj_uw kimmonismus officiallogank designarena alezander907 giffmana jeremyphoward hamelhusain
AI News for 5/4/2026-5/5/2026 highlights a shift in AI product development emphasizing model + harness + workflow + UI + memory + economics over model quality alone, with notable updates from OpenAI Codex and Claude including new features like Appshots, auto mode, and Sonnet 4.6. DeepSeek made a significant market impact by permanently discounting DeepSeek-V4-Pro by 75%, drastically improving cost/performance ratios compared to Gemini 3.1 Pro, GPT-5.5, and Claude Opus 4.7. Meanwhile, Gemini 3.5 Flash showed benchmark improvements but received mixed feedback on practical utility. The competitive landscape continues to tighten with Qwen and other Chinese frontier models.
not much happened today
vllm-0.20.0 poolside-laguna-xs.2 ling-2.6-flash nemotron-3-nano-omni qwen-3.5 vllm poolside nvidia opensrouter lmstudio ollama unsloth fal fireworks deepinfra togethercompute baseten canonical memory-optimization mixture-of-experts model-optimization inference-speed quantization model-deployment multimodality hardware-optimization model-benchmarking open-models agentic-ai jeremyphoward maharshii teortaxestex aymericroucher piotrz
vLLM v0.20.0 introduces significant improvements in memory and MoE serving efficiency, including TurboQuant 2-bit KV cache for 4× KV capacity and a 2.1% latency improvement. The update supports multiple hardware platforms like DeepSeek V4 MegaMoE on Blackwell, Jetson Thor, ROCm, Intel XPU, and Grace-Blackwell setups. Early benchmarks show DeepSeek V4 Pro on B300 hardware can be up to 8× faster than H200. The ecosystem is rapidly adopting day-0 support for new open models such as Poolside Laguna XS.2, Ling-2.6-flash, and NVIDIA Nemotron 3 Nano Omni.
Poolside released Laguna XS.2, a 33B total / 3B active MoE coding model under Apache 2.0, capable of running on a single GPU, with hybrid attention and FP8 KV cache, performing near Qwen-3.5.
NVIDIA launched Nemotron 3 Nano Omni, a 30B / A3B multimodal MoE with 256K context, supporting text, image, video, audio, and documents, with immediate distribution across multiple platforms. Discussions highlighted tradeoffs in quantization methods and a shift away from CUDA lock-in towards heterogeneous accelerator support.
Gemma 4
gemma-4 gemma-4-31b gemma-4-26b-a4b google-deepmind multimodality long-context model-architecture moe local-inference model-optimization function-calling quantization jeffdean _philschmid rasbt ggerganov clattner_llvm julien_c clementdelangue
Google DeepMind released Gemma 4, a family of open-weight, multimodal models with long-context support up to 256K tokens under an Apache 2.0 license, marking a major capability and licensing shift. The lineup includes 31B dense, 26B MoE (A4B), and two edge models (E4B, E2B) optimized for local and edge deployment with native multimodal support (text, vision, audio). Early benchmarks show Gemma-4-31B ranking #3 among open models and strong scientific reasoning performance with 85.7% GPQA Diamond. Day-0 ecosystem support includes llama.cpp, Ollama, vLLM, and LM Studio, with notable local inference performance on hardware like M2 Ultra and RTX 4090. The architecture features hybrid attention and MoE layering, diverging from standard transformers. Community and developer engagement is high, with rapid adoption and tooling integration.
not much happened today
nemotron-3-super gpt-oss-120b qwen3.5-122b-a10b nvidia perplexity replit base44 vllm llama.cpp ollama togethercompute baseten wandb langchain unsloth model-architecture model-optimization inference-speed kv-cache multi-token-prediction agent-infrastructure orchestration persistent-agents model-serving product-launches karpathy ctnzr bnjmn_marie artificialanlys
NVIDIA’s Nemotron 3 Super is a 120B parameter / ~12B active open model featuring a hybrid Mamba-Transformer / SSM Latent MoE architecture and 1M context window, delivering up to 2.2x faster inference than GPT-OSS-120B in FP4 with strong throughput gains. It supports agentic workloads and is unusually open with weights, data, and infrastructure details released. The model scored 36 on the AA Intelligence Index, outperforming GPT-OSS-120B but behind Qwen3.5-122B-A10B. Community and infrastructure support from projects like vLLM, llama.cpp, Ollama, Together, Baseten, W&B Inference, LangChain, and Unsloth GGUFs was immediate. Key technical innovations include native multi-token prediction (MTP) and a significant KV-cache efficiency advantage.
On the product side, a shift towards persistent agent runtimes and orchestration layers is highlighted, with Andrej Karpathy advocating for a "bigger IDE" concept where agents replace files as the unit of work, enabling legible, forkable agentic organizations with real-time control. New launches fitting this vision include Perplexity’s Personal Computer, an always-on local/cloud hybrid running on Mac mini, and Computer for Enterprise orchestrating 20 specialized models and 400+ apps. Replit Agent 4 offers a collaborative, canvas-like workflow with parallel agents, while Base44 Superagents provide integrated solutions for nontechnical users. The engineering focus is increasingly on the orchestration harness rather than just the model.
Context Graphs: Hype or actually Trillion-dollar opportunity?
glm-ocr qwen3-coder-next sera-14b zhipu-ai lmsys vllm novita-labs ollama alibaba allenai cognition cursor multimodality ocr long-context mixture-of-experts agentic-coding-models context-graphs benchmarking model-deployment model-optimization model-training jaya_gupta dharmesh_shah
Zhipu AI launched GLM-OCR, a lightweight 0.9B multimodal OCR model excelling in complex document understanding with top benchmark scores and day-0 deployment support from lmsys, vllm, and novita labs. Ollama enabled local-first usage with easy offline operation. Alibaba released Qwen3-Coder-Next, an 80B MoE model with only 3B active parameters, designed for coding agents with a massive 256K context window and trained on 800K verifiable tasks, achieving over 70% SWE-Bench Verified. The open coding ecosystem also saw Allen AI announce SERA-14B, an on-device-friendly coding model with new datasets. The emerging concept of Context Graphs was highlighted as a promising framework for data and agent traceability, with initiatives like Cursor's Agent Trace specifying context graphs for coding agents, emphasizing potential improvements in agent performance and customer-driven adoption. This coverage reflects ongoing innovation in multimodality, long-context, mixture-of-experts, and agentic coding models.
not much happened today
gpt-5.2 claude-opus-4.5 kimi-k2.5 openai anthropic deeplearningai langchain apple agentic-ai multimodality coding self-verification agent-engineering model-benchmarking model-optimization workflow-automation
AI News for 1/27/2026-1/28/2026 highlights a quiet day with deep dives into frontier model "personality split" where GPT-5.2 excels at exploration and Claude Opus 4.5 at exploitation, suggesting OpenAI suits research workflows and Anthropic commercial reliability. The rise of agentic coding loops shows new failure modes, with self-verification workflows gaining traction. The open-model Kimi K2.5 emerges as a flashpoint, boasting enhanced agent execution, multimodality, and coding polish, runnable on Apple silicon M3 Ultra Mac Studios with Thunderbolt 5 (RDMA), and challenging Claude Opus 4.5 on benchmarks and pricing. Licensing issues threaten enterprise adoption despite model quality. The meme "clawdbot" reflects rapid agent branding proliferation. Agent engineering advances with shared "skills" interfaces promoted by DeepLearning.AI, Anthropic, and LangChain.
Moonshot Kimi K2.5 - Beats Sonnet 4.5 at half the cost, SOTA Open Model, first Native Image+Video, 100 parallel Agent Swarm manager
kimi-k2.5 moonshotai multimodality model-training mixture-of-experts agentic-ai vision video-understanding model-optimization parallel-processing office-productivity
MoonshotAI's Kimi K2.5 is a 32B active-1T parameter open-weights model featuring native multimodality with image and video understanding, built through continual pretraining on 15 trillion mixed visual and text tokens. It introduces a new MoonViT vision encoder and supports advanced capabilities like Agent Swarm, which coordinates up to 100 sub-agents for parallel workflows, and an Office Productivity K2.5 Agent for large-scale office tasks. This release marks a significant leap in open models from China, claiming state-of-the-art results on benchmarks like HLE and BrowseComp, and offering aggressive API pricing and throughput.
not much happened today
glm-4.7-flash grok deepseek-r1 qwq x-ai unsloth-ai google deepseek ollama transformer-architecture recommendation-systems local-inference kv-cache quantization tensor-parallelism reasoning model-optimization fine-tuning giffmana david_sholz yuchenj_uw nearcyan sam_paech teortaxes_tex danielhanchen alexocheema nopmobiel rohanpaul_ai
X Engineering open-sourced its new transformer-based recommender algorithm, sparking community debate on transparency and fairness. GLM-4.7-Flash (30B-A3B) gains momentum as a strong local inference model with efficient KV-cache management and quantization tuning strategies. Innovations include tensor parallelism on Mac Minis achieving ~100 tok/s throughput. Research highlights "Societies of Thought" as a reasoning mechanism improving model accuracy by 20%+.
Apple picks Google's Gemini to power Siri's next generation
gemini claude chatgpt engram apple google openai anthropic deepseek conditional-memory long-context hashing memory-optimization transformers model-scaling sparsity hardware-optimization model-architecture ai-healthcare model-optimization
Apple has decided to power Siri with Google's Gemini models and cloud technology, marking a significant partnership and a setback for OpenAI, which was initially partnered with Apple. Anthropic launched "Cowork," a product preview for Claude's coding capabilities, sparking discussions about "LLM OS". OpenAI introduced ChatGPT Health and acquired Torch to expand in healthcare AI. DeepSeek unveiled Engram, a new conditional memory module that enables O(1) lookup-style memory for static patterns, improving long-context handling and offering hardware-friendly optimizations to scale knowledge capacity efficiently. Engram is positioned as a key modeling primitive for next-gen sparse models, with ongoing community debate about its architectural merits and practical impact.
not much happened today
glm-4.7 glm-4.6 minimax-m2.1 gemma-3 gemma-scope-2 google-deepmind valsai minimax-ai ollama trae alibaba sophont prime-intellect interpretability sparse-autoencoders agent-workflows model-benchmarking medical-evaluation multi-agent-systems model-performance model-optimization reinforcement-learning tool-use function-calling context-windows ivanfioravanti awnihannun deedydas cline omarsar0 adonis_singh eliebakouch teortaxestex ibragim_bad callum_mcdougall neelnanda5
GLM-4.7 and MiniMax M2.1 open-weight model releases highlight day-0 ecosystem support, coding throughput, and agent workflows, with GLM-4.7 achieving a +9.5% improvement over GLM-4.6 and MiniMax M2.1 positioned as an OSS Claude-like MoE model with 230B total parameters and 200K context. Gemma Scope 2 from google-deepmind introduces sparse autoencoders and transcoders for interpretability across Gemma 3 models, aiming to provide shared infrastructure for safety and debugging. The Medmarks v0.1 open medical evaluation suite and leaderboard launch addresses the need for open medical benchmarking across 15+ environments, engaging clinicians and researchers.
OpenAI GPT Image-1.5 claims to beat Nano Banana Pro, #1 across all Arenas, but completely fails Vibe Checks
gpt-image-1.5 nano-banana-pro mimo-v2-flash deepseek-v3.2 openai gemini xiaomi lmsys deepseek openrouter image-generation instruction-following benchmarking model-efficiency long-context multi-token-prediction hybrid-attention model-optimization inference-speed agentic-workflows model-architecture model-quantization fuli_luo eliebakouch
OpenAI released its new image model GPT Image 1.5, featuring precise image editing, better instruction following, improved text and markdown rendering, and faster generation up to 4×. Despite topping multiple leaderboards like LMArena (1277), Design Arena (1344), and AA Arena (1272), user feedback from Twitter, Reddit, and Discord communities is largely negative compared to Nano Banana Pro by Gemini. Xiaomi introduced the MiMo-V2-Flash, a 309B MoE model optimized for inference efficiency with 256K context window, achieving state-of-the-art scores on SWE-Bench. The model uses Hybrid Sliding Window Attention and multi-token prediction, offering significant speedups and efficiency improvements. The timing of OpenAI's launch amid competition from Gemini and Nano Banana Pro affects user sentiment, highlighting challenges in benchmarking relevance.
NVIDIA Nemotron 3: hybrid Mamba-Transformer completely open source models from 30B to 500B
nemotron-3-nano qwen3-30b-a3b-base nvidia huggingface togethercompute baseten vllm llamaindex hybrid-architecture mixture-of-experts reinforcement-learning long-context model-release open-source-models model-training model-optimization benchmarking agent-training ctnzr andrew_n_carr awnihannun
NVIDIA has released Nemotron 3 Nano, a fully open-source hybrid Mamba-Transformer Mixture-of-Experts (MoE) model with a 30B parameter size and a 1 million token context window. It includes open weights, training recipes, datasets, and an RL environment suite called NeMo Gym, supporting commercial use under the NVIDIA Open Model License. The model achieves state-of-the-art results on benchmarks like SWE-Bench and Artificial Analysis Intelligence Index, outperforming Qwen3-30B A3B. Ecosystem support is immediate with integrations into inference stacks like vLLM, llama.cpp, and Baseten. Upcoming larger models, Nemotron Super and Ultra, will feature NVFP4 pretraining and LatentMoE routing to optimize compute. This release marks a significant milestone for open-source American AI with comprehensive open assets and advanced hybrid architecture.
not much happened today
gpt-5.2 opus-4.5 gemini-3-pro gpt-5.1 olmo-3.1-32b qwen3-vl-235b openai allen_ai mistral-ai ollama lmstudio thinkymachines reinforcement-learning model-benchmarking long-context model-quantization model-optimization inference-speed sparsity fine-tuning vision sama scaling01 akhaliq artificialanlys lechmazur acerfur epochairesearch
GPT-5.2 shows mixed performance in public evaluations, excelling in agentic tasks but at a significantly higher cost (~$620/run) compared to Opus 4.5 and GPT-5.1. It performs variably on reasoning and coding benchmarks, with some improvements on long-context tasks. Extended "reasoning effort" settings notably impact results. Aggregators rank Gemini 3 Pro above GPT-5.2 in task persistence. OpenAI released sparse activation models sparking debate on sparsity vs MoE architectures. Allen AI's Olmo 3.1 (32B) advances open reinforcement learning scale with substantial compute investment (~125k H100 hours). Mistral's Devstral-2 and llama.cpp improve local inference infrastructure with new features like GGUF support and distributed speedups. Tinker platform goes GA with vision input and finetuning support for Qwen3-VL-235B.
MCP -> Agentic AI Foundation, Mistral Devstral 2
devstral-2 devstral-small-2 sonnet-4.3 deepseek-v3.2 qwen3-vl openai anthropic block mistral-ai alibaba linux-foundation deepseek agentic-ai coding-models reinforcement-learning model-performance model-optimization open-weights cli-tools multi-file-code-automation data-decontamination moe reward-models rl-stability guillaumelample b_roziere qtnx_ charliermarsh omarsar0 eliebakouch justinwaugh cwolferesearch pan
OpenAI Engineering sees a significant collaborative milestone with the launch of the Agentic AI Foundation under the Linux Foundation, uniting projects from Anthropic, OpenAI, and Block. Mistral released Devstral 2, a coding model with 123B parameters and open weights, offering a cost-effective alternative to Sonnet 4.3 and competitive performance against DeepSeek v3.2. The new Mistral Vibe CLI supports agentic coding workflows with rapid ecosystem integration. Alibaba introduced Soft Adaptive Policy Optimization (SAPO) for reinforcement learning tuning, improving stability and performance in Qwen3-VL across multiple tasks. Research highlights include the importance of data decontamination in RL and ongoing discussions on MoE RL stability and reward hacking mitigation.
not much happened today
glm-4.6v glm-4.6v-flash jina-vlm-2b hugging-face zhipu-ai jina-ai google-deepmind axiomprover fine-tuning multimodality model-optimization long-context mechanistic-interpretability formal-methods sequence-architectures reinforcement-learning lioronai akshay_pachaar _akhaliq ben_burtenshaw vllm_project prince_canuma zenmuxai eliebakouch theturingpost axiommathai neelnanda5 sarahookr
Claude Code Skills gains attention with a published talk and Hugging Face's new "skill" enabling one-line fine-tuning pipelines for models from ~0.5B to 70B parameters, supporting SFT, DPO, and GRPO, costing as low as ~$0.30 for small runs. Zhipu AI launches multimodal models GLM-4.6V (106B params MoE) and GLM-4.6V-Flash (9B dense), featuring 128k context and native multimodal function calling, with free Flash variant and API pricing detailed. Jina AI releases Jina-VLM (2B), a compact multilingual VLM excelling in diagrams and documents with top benchmark scores. At NeurIPS 2025, research highlights include Google's post-Transformer sequence architectures (Moneta, Yaad, Memora) showing up to 20% gains in long-context retrieval, AxiomProver's autonomous Lean system solving 9/12 Putnam 2025 problems rapidly, and mechanistic interpretability advances discussed by Chris Olah emphasizing scalable tooling.
not much happened today
vllm-0.12.0 gemma3n qwen3-omni qwen3-vl gpt-5.1-codex-max gemini-3-pro runway-gen-4.5 kling-video-2.6 vllm nvidia huggingface langchain-ai together-ai meta-ai-fair sonarsource openrouter runway gemini arena gpu-programming quantization multimodality agent-platforms reinforcement-learning static-analysis reasoning inference-infrastructure model-optimization economics audio video-generation jeremyphoward mervenoyann sydneyrunkle swyx maximelabonne
vLLM 0.12.0 introduces DeepSeek support, GPU Model Runner V2, and quantization improvements with PyTorch 2.9.0 and CUDA 12.9. NVIDIA launches CUDA Tile IR and cuTile Python for advanced GPU tensor operations targeting Blackwell GPUs. Hugging Face releases Transformers v5 RC with an any-to-any multimodal pipeline supporting models like Gemma3n and Qwen3-Omni. Agent platforms see updates from LangChain with content moderation and cost tracking, Together AI and Meta AI collaborate on RL for long-horizon workflows, and SonarSource integrates static analysis into AI codegen. Economic insights from OpenRouter highlight coding as a key AI application, with reasoning models surpassing 50% usage and market bifurcation between premium and open models. Additionally, Kling Video 2.6 debuts native audio capabilities, and Runway Gen-4.5, Qwen3-TTS, and Gemini 3 Pro advance multimodality.
not much happened today
claude-opus-4.5 qwen-3-4b qwen-3-8b qwen-3-14b deepseek-r1 anthropic booking.com perplexity-ai langchain claude scaling01 deepseek qwen prefect agent-systems multi-agent-systems reasoning benchmarking cost-efficiency model-optimization long-context memory-management reinforcement-learning model-performance multi-agent-communication latent-representation inference-cost software-integration jeremyphoward alexalbert__ omarsar0 lingyang_pu dair_ai
Anthropic introduces durable agents and MCP tasks for long-running workflows, with practical engineering patterns and integrations like Prefect. Booking.com deploys a large-scale agent system improving customer satisfaction using LangGraph, Kubernetes, GPT-4 Mini, and Weaviate. Perplexity rolls out user-level memory and virtual try-on features. Claude Opus 4.5 leads on LisanBench and Code Arena WebDev benchmarks with mixed community feedback on its "thinking" and "non-thinking" modes, while improving cost-efficiency and UX with batch APIs and context compaction. Research on multi-agent systems shows LatentMAS reduces communication tokens by 70-84% and improves accuracy using Qwen3 models, and reasoning trace distillation achieves significant token reduction with maintained accuracy, highlighting the importance of reasoning trace style.
not much happened today
kimi-k2-thinking kimi-k3 gelato-30b-a3b omnilingual-wav2vec-2.0 moonshot-ai meta-ai-fair togethercompute qwen attention-mechanisms quantization fine-tuning model-optimization agentic-ai speech-recognition multilingual-models gui-manipulation image-editing dataset-release yuchenj_uw scaling01 code_star omarsar0 kimi_moonshot anas_awadalla akhaliq minchoi
Moonshot AI's Kimi K2 Thinking AMA revealed a hybrid attention stack using KDA + NoPE MLA outperforming full MLA + RoPE, with the Muon optimizer scaling to ~1T parameters and native INT4 QAT for cost-efficient inference. K2 Thinking ranks highly on LisanBench and LM Arena Text leaderboards, offering low-cost INT4 serving and strong performance in Math, Coding, and Creative Writing. It supports heavy agentic tool use with up to 300 tool requests per run and recommends using the official API for reliable long-trace inference. Meta AI released the Omnilingual ASR suite covering 1600+ languages including 500 underserved, plus a 7B wav2vec 2.0 model and ASR corpus. Additionally, the Gelato-30B-A3B model for computer grounding in GUI manipulation agents outperforms larger VLMs, targeting immediate agent gains. Qwen's image-edit LoRAs and light-restoration app were also highlighted.
Terminal-Bench 2.0 and Harbor
kimi-k2-thinking moonshot-ai anthropic hugging-face ollama slime-framework benchmarking agentic-ai quantization model-optimization inference model-deployment moe context-windows cost-efficiency clementdelangue dbreunig awnihannun crystalsssup kimi_moonshot
Terminal-Bench has fixed task issues and launched version 2.0 with cloud container support via the Harbor framework, gaining recognition from models like Claude 4.5 and Kimi K2 Thinking. Moonshot AI's Kimi K2 Thinking is a 1 trillion parameter MoE reasoning model with ~32B active parameters, running natively in INT4 quantization and featuring a 256K context window. It leads open-weights benchmarks with an Artificial Analysis Intelligence Index score of 67 and strong agentic performance, running efficiently on consumer Apple silicon and 2× M3 Ultra hardware. The model is broadly available on Hugging Face, Ollama Cloud, and integrated into frameworks like slime. Serving bottlenecks were traced to network bandwidth rather than GPU limits, highlighting infrastructure considerations for LLM deployment.
Cursor 2.0 & Composer-1: Fast Models and New Agents UI
composer-1 gpt-oss-safeguard-20b gpt-oss-safeguard-120b gpt-oss gpt-5-mini cursor_ai openai huggingface ollama cerebras groq goodfireai rakuten agentic-coding reinforcement-learning mixture-of-experts fine-tuning policy-classification open-weight-models inference-stacks cost-efficiency multi-agent-systems ide voice-to-code code-review built-in-browser model-optimization sasha_rush dan_shipper samkottler ellev3n11 swyx
Cursor 2.0 launched with Composer-1, an agentic coding model optimized for speed and precision, featuring multi-agent orchestration, built-in browser for testing, and voice-to-code capabilities. OpenAI released gpt-oss-safeguard models (20B, 120B) for policy-based safety classification, open-weight and fine-tuned from gpt-oss, available on Hugging Face and supported by inference stacks like Ollama and Cerebras. Goodfire and Rakuten demonstrated sparse autoencoders for PII detection matching gpt-5-mini accuracy at significantly lower cost. The Cursor 2.0 update also includes a redesigned interface for managing multiple AI coding agents, marking a major advancement in AI IDEs. "Fast-not-slowest" tradeoff emphasized by early users for Composer-1, enabling rapid iteration with human-in-the-loop.
not much happened today
nemotron-nano-2 gpt-oss-120b qwen3 llama-3 minimax-m2 glm-4.6-air gemini-2.5-flash gpt-5.1-mini tahoe-x1 vllm_project nvidia mistral-ai baseten huggingface thinking-machines deeplearningai pytorch arena yupp-ai zhipu-ai scaling01 stanford transformer-architecture model-optimization inference distributed-training multi-gpu-support performance-optimization agents observability model-evaluation reinforcement-learning model-provenance statistical-testing foundation-models cancer-biology model-fine-tuning swyx dvilasuero _lewtun clementdelangue zephyr_z9 skylermiao7 teortaxestex nalidoust
vLLM announced support for NVIDIA Nemotron Nano 2, featuring a hybrid Transformer–Mamba design and tunable "thinking budget" enabling up to 6× faster token generation. Mistral AI Studio launched a production platform for agents with deep observability. Baseten reported high throughput (650 TPS) for GPT-OSS 120B on NVIDIA hardware. Hugging Face InspectAI added inference provider integration for cross-provider evaluation. Thinking Machines Tinker abstracts distributed fine-tuning for open-weight LLMs like Qwen3 and Llama 3. In China, MiniMax M2 shows competitive performance with top models and is optimized for agents and coding, while Zhipu GLM-4.6-Air focuses on reliability and scaling for coding tasks. Rumors suggest Gemini 2.5 Flash may be a >500B parameter MoE model, and a possible GPT-5.1 mini reference appeared. Outside LLMs, Tahoe-x1 (3B) foundation model achieved SOTA in cancer cell biology benchmarks. Research from Stanford introduces a method to detect model provenance via training-order "palimpsest" with strong statistical guarantees.
not much happened today
qwen3-vl-4b qwen3-vl-8b qwen2.5-vl-72b deepseek-v3.1 alibaba arena runway nvidia togethercompute ollama model-optimization fine-tuning inference-speed video-generation diffusion-models representation-learning local-ai speculative-decoding fp8-quantization context-windows karpathy
Alibaba released compact dense Qwen3-VL models at 4B and 8B sizes with FP8 options, supporting up to 1M context and open vocabulary detection, rivaling larger models like Qwen2.5-VL-72B. Ecosystem support includes MLX-VLM, LM Studio, vLLM, Kaggle models, and Ollama Cloud. In video AI, Arena added Sora 2 models leading in video benchmarks, with Higgsfield Enhancer improving video quality. Runway launched domain-specific workflow apps for creative tasks. Research on Representation Autoencoders for DiTs (RAE-DiT) shows improved diffusion model performance. On local training, NVIDIA DGX Spark enables strong local fine-tuning, while Nanochat by Karpathy offers a minimal stack for training and inference. Together AI introduced ATLAS, a speculative decoding method achieving up to 4× faster inference on DeepSeek-V3.1. These developments highlight advances in efficient model deployment, video AI, local fine-tuning, and inference speed optimization.
Thinking Machines' Tinker: LoRA based LLM fine-tuning API
qwen-235b-a22b sora-2 thinking-machines openai fine-tuning lora model-training api model-optimization distributed-training post-training-methods research-productivity video-generation content-moderation engagement-patterns karpathy lilianweng sama
Thinking Machines recently raised $2 billion without shipping a product until now, launching their first product Tinker, a managed service API for fine-tuning large and mixture-of-experts models like Qwen-235B-A22B using LoRA for cost-efficient training. The Tinker API offers low-level primitives for post-training methods and is supported by an open-source Tinker Cookbook library. Influential AI figures like Andrej Karpathy and Lilian Weng praised its design for reducing complexity and boosting research productivity. Meanwhile, OpenAI launched Sora 2, a video+audio model integrated into their consumer social app, sparking viral engagement and concerns over misuse and content moderation. Sam Altman emphasized the product's dual focus on delight and revenue alongside AGI research.
not much happened today
gemini-robotics-1.5 gemini-live embeddinggemma veo-3 gemini-2.5-flash code-world-model-32b qwen3-coder-30b vllm-v1 mlx-lm flashattention-4 google meta-ai-fair perplexity-ai baseten spatial-reasoning temporal-reasoning agentic-ai code-semantics code-execution-traces coding-infrastructure runtime-optimization batch-inference embedding-latency api model-optimization model-performance osanseviero _anniexie rmstein scaling01 giffmana cline redhat_ai awnihannun charles_irl bernhardsson akshat_b aravsrinivas
Google released a dense September update including Gemini Robotics 1.5 with enhanced spatial/temporal reasoning, Gemini Live, EmbeddingGemma, and Veo 3 GA powering creative workflows. They also introduced agentic features like restaurant-reservation agents and reduced pricing for Gemini 2.5 Flash. Meta AI unveiled the open-weight Code World Model (CWM) 32B, excelling in code semantics and math benchmarks, with innovations in training code models via execution traces. Local-first coding setups highlight Qwen3-Coder-30B running efficiently on consumer GPUs, paired with tools like Cline and LM Studio. Runtime improvements include vLLM v1 supporting hybrid models and mlx-lm adding batch inference on Apple silicon. In infrastructure, FlashAttention 4 was reverse-engineered revealing a ~20% speedup from architectural optimizations. Perplexity AI advances its independent web index and browsing API with upcoming feed refreshes. Embedding latency improvements were achieved by Superhuman using Baseten.
not much happened today
qwen3-max qwen3-vl qwen3-coder-plus gpt-5-codex code-world-model-32b claude-sonnet-4 claude-opus-4.1 alibaba openai meta-ai-fair huggingface anthropic microsoft github context-windows code-generation model-releases model-benchmarking api model-optimization multimodality software-engineering model-training huybery akhaliq lmarena_ai gdb ylecun pierceboggan julesagent
Alibaba unveiled the Qwen3 model family including Qwen3-Max and Qwen3-VL with a native 256K context window expandable to 1M, strong OCR in 32 languages, and rapid release velocity (~3.5 releases/month) backed by a $52B infrastructure roadmap. OpenAI launched GPT-5 Codex, an agent-optimized coding model with up to 400K context and adaptive reasoning priced at $1.25/$10 per million tokens, integrated into Cline and benchmarked in WebDev arenas. Meta AI FAIR released the open-weight Code World Model (CWM) 32B, a dense code generation model with strong benchmark scores (e.g., 65.8% SWE-bench Verified, 96.6% Math-500) and public safety reports. Ecosystem updates include GitHub Copilot's new embedding model for faster code search and Anthropic's Claude Sonnet 4 and Opus 4.1 integration into Microsoft 365 Copilot. The vLLM 0.10.2 update introduces Decode Context Parallel (DCP) for improved system performance.
Grok 4 Fast: Xai's distilled, 40% more token efficient, 2m context, 344 tok/s frontier model
grok-4-fast magistral-1.2 moondream-3 granite-docling-258m sail-vl2 xai meta-ai-fair mistral-ai ibm bytedance efficiency reasoning vision multimodality model-optimization model-deployment vision-encoders model-architecture model-training nearcyan aidangomez _akhaliq vikhyatk rohanpaul_ai
xAI announced Grok 4 Fast, a highly efficient model running at 344 tokens/second, offering reasoning and nonreasoning modes and free trials on major platforms. Meta showcased its neural band and Ray-Ban Display with a live demo that experienced hiccups but sparked discussion on live hardware demos and integration challenges. Meta is also developing a first-party "Horizon Engine" for AI rendering and released Quest-native Gaussian Splatting capture tech. New model releases include Mistral's Magistral 1.2, a compact multimodal vision-language model with improved benchmarks and local deployment; Moondream 3, a 9B-parameter MoE VLM focused on efficient visual reasoning; IBM's Granite-Docling-258M, a document VLM for layout-faithful PDF to HTML/Markdown conversion; and ByteDance's SAIL-VL2, a vision-language foundation model excelling at multimodal understanding and reasoning at 2B and 8B parameter scales.
Softbank, NVIDIA and US Govt take 2%, 5% and 10% of Intel, will develop Intel x86 RTX SOCs for consumer & datacenters
magistral-1.2 moondream-3 nvidia intel meta-ai-fair mistral-ai multimodality vision model-optimization model-efficiency model-architecture reinforcement-learning fine-tuning ai-hardware gaussian-splatting live-demo visual-reasoning nearcyan _akhaliq vikhyatk
Nvidia and Intel announced a joint development partnership for multiple new generations of x86 products, marking a significant shift in the tech industry. This collaboration has been in the works for a year and impacts both consumer and data center markets, boosting hopes for Intel's Foundry business. On the AI hardware front, Meta showcased its neural band and Ray-Ban Display with a live demo that experienced hiccups but sparked discussion on live tech demos. Meta is also moving from Unity to its own Horizon Engine for AI rendering, including Gaussian splatting capture technology. In AI models, Mistral released Magistral 1.2, a compact multimodal vision-language model with improved benchmarks and local deployment capabilities, while Moondream 3 previewed a 9B-parameter, 2B-active MoE VLM focused on efficient visual reasoning.
not much happened today
gpt-5-codex vllm-0.10.2 qwen3-next-80b hunyuanimage-2.1 openai microsoft perplexity-ai huggingface amd tencent lmstudio agentic-ai ide context-windows inference distributed-inference reinforcement-learning robotics long-context model-optimization text-to-image multimodality model-licenses gdb teknium1 finbarrtimbers thsottiaux theturingpost pierceboggan amandaksilver aravsrinivas sergiopaniego art_zucker danielhanchen rwojo awnihannun
GPT-5 Codex rollout shows strong agentic coding capabilities with some token bloat issues. IDEs like VS Code Insiders and Cursor 1.6 enhance context windows and model integration. vLLM 0.10.2 supports aarch64 and NVIDIA GB200 with performance improvements. AMD ROCm updates add modern attention, sparse MoE, and distributed inference. TRL introduces Context Parallelism for long-context training. Robotics and RL data pipelines improve with Unsloth and LeRobotDataset v3. Qwen3-Next-80B runs efficiently on Mac M4 Max with MLX. Tencent's HunyuanImage 2.1 is a 17B bilingual text-to-image model with 2048×2048 resolution and restricted open weights.
GPT-5 Codex launch and OpenAI's quiet rise in Agentic Coding
gpt-5-codex qwen3-next-80b openai alibaba together-ai nvidia agentic-ai software-engineering long-context mixture-of-experts model-optimization cuda-acceleration inference-efficiency routing task-adaptive-thinking sama swyx omarsar0 ofirpress
OpenAI released GPT-5-Codex, an agentic coding model optimized for long-running software engineering tasks with dynamic task-adaptive thinking, multi-hour autonomy, and improved code quality. It achieves 51% accuracy on an unreleased large refactor benchmark and integrates deeply with developer tools like Xcode. Meanwhile, Alibaba launched Qwen3-Next-80B, a hybrid MoE model with native long-context support (262k tokens, extensible to 1M+), targeting efficient reasoning and repository-scale code analysis, supported by Together AI and NVIDIA with CUDA-accelerated attention. The trend towards hybrid SSM + MoE architectures is noted, emphasizing efficiency and scaling in China and US training regimes. Community discussions highlight the importance of variable compute and routing for inference efficiency and quality.
Oracle jumps +36% in a day after winning $300B OpenAI contract
qwen3-235b qwen3-4b qwen2.5-7b vllm oracle openai microsoft moonshot-ai vllm-project thinking-machines-lab meta reinforcement-learning model-weight-updates deterministic-inference benchmarking long-context model-optimization cuda distributed-training kimi_moonshot arankomatsuzaki qgallouedec cHHillee woosuk_k stasbekman
Oracle's OCI division reported a stunning +359% revenue bookings growth to $455B with cloud revenue guidance of $144B by 2030, driven significantly by a large deal with OpenAI amid tensions with Microsoft. On AI infrastructure, Moonshot AI released Kimi’s checkpoint-engine, enabling rapid weight updates on 1T-parameter models across thousands of GPUs, integrating with vLLM. RLFactory introduced a plug-and-play reinforcement learning framework for tool-using agents, showing smaller models outperforming larger ones. TRL v0.23 added context parallelism for long-context training. Thinking Machines Lab published research on deterministic inference pipelines, making vLLM deterministic for Qwen models. Meta launched BackendBench, a PyTorch benchmarking tool.
not much happened today
gpt-5 kimi-k2-0905 glm-4.5 qwen3-asr opus-4.1 cognition founders-fund lux-capital 8vc neo vercel claude groq alibaba huggingface meta-ai-fair google theturingpost algoperf coding-agents agent-architecture open-source model-evaluation multilingual-models speech-recognition model-optimization kv-cache quantization algorithmic-benchmarking video-generation context-windows swyx tim_dettmers
Cognition raised $400M at a $10.2B valuation to advance AI coding agents, with swyx joining to support the "Decade of Agents" thesis. Vercel launched an OSS "vibe coding platform" using a tuned GPT-5 agent loop. Claude Code emphasizes minimalism in agent loops for reliability. Kimi K2-0905 achieved 94% on coding evals and improved agentic capabilities with doubled context length. Alibaba released Qwen3-ASR, a multilingual transcription model with <8% WER. Meta introduced Set Block Decoding for 3-5× faster decoding without architectural changes. Innovations in KV cache compression and quantization include AutoRound, QuTLASS v0.1.0, and AlgoPerf v0.6. Google's Veo 3 video generation API went GA with significant price cuts and vertical video support.
not much happened today
gpt-5 gpt-5-high gpt-5-mini-high gpt-5-nano-high imagen-4 gemma-3-270m openai google lmsys model-releases model-performance prompt-engineering developer-tools image-generation model-optimization transformers tokenization model-scaling sama aidan_mclau kevinweil lmarena_ai edwinarbus gdb omarsar0 philschmid m4rkmc
OpenAI rolled out GPT-5 as the default in ChatGPT with new modes and a "warmer" personality, plus expanded message limits for Plus/Team users and Enterprise/Edu access. Performance rankings show gpt-5-high leading, with smaller variants also ranked, though critiques note some underperformance versus Chinese models and sensitivity to sycophancy. OpenAI enhanced developer tools with a "Quick eval" feature, coding tips, and an improved Playground. Google released Imagen 4 generally available with faster generation and higher resolution, plus the ultra-small Gemma 3 270M model with a large vocabulary and ecosystem support. Podcasts featured OpenAI leaders discussing GPT-5 systems, routing, and efficiency.
Western Open Models get Funding: Cohere $500m @ 6.8B, AI2 gets $152m NSF+NVIDIA grants
gpt-5 o3 command-a gemma-3-270m imagen-4 dinov3 openai perplexity-ai ai2 nvidia cohere meta-ai-fair google hugging-face ollama unsloth model-speed funding ai-infrastructure on-device-ai quantization embedding-models image-generation self-supervised-learning vision dense-prediction benchmarking instruction-following model-optimization model-release challenge joelle_pineau fchollet awnihannun _philschmid osanseviero
OpenAI's GPT-5 achieved a speedrun of Pokemon Red 3x faster than o3. Perplexity raised $200M at a $20B valuation. AI2 secured $75M NSF grants and $77M from NVIDIA for AI infrastructure projects like Olmo and Molmo. Cohere raised $500M and hired Joelle Pineau from meta-ai-fair, boosting models like Command A. Google released the Gemma 3 270M on-device tiny LLM with INT4 QAT checkpoints and large embedding tables, and made Imagen 4 generally available with a fast version at $0.02/image. Meta-ai-fair introduced DINOv3, a family of self-supervised vision foundation models with high-resolution dense features and strong performance on benchmarks like COCO detection and ADE20K segmentation, under a permissive license. A $150,000 MiniMax AI Agent Challenge is ongoing with 200+ prizes, encouraging AI project builds by August 25.
not much happened today
gpt-5 gpt-oss-120b opus-4.1 sonnet-4 openai anthropic minimax context-windows model-routing model-hosting multi-tool-pipelines prompt-caching model-extraction model-pairing cost-efficiency model-optimization sama jeremyphoward jxmnop _catwu
OpenAI continues small updates to GPT-5, introducing "Auto/Fast/Thinking" modes with 196k token context, 3,000 messages/week, and dynamic routing to cheaper models for cost efficiency. The MiniMax AI Agent Challenge offers $150,000 in prizes for AI agent development by August 25. The community discusses GPT-OSS-120B base model extraction, hosting, and tooling improvements, including multi-tool pipelines and flex-attention. Anthropic announces model pairing in Claude Code with Opus 4.1 for planning and Sonnet 4 for execution, expanding context to 1M tokens and introducing prompt caching. Key figures include @sama, @jeremyphoward, @jxmnop, and @_catwu.
Gemini 2.5 Deep Think finally ships
gemini-2.5-deep-think gpt-oss gpt-5 kimi-k2-turbo-preview qwen3-coder-flash glm-4.5 step-3 claude openai anthropic google-deepmind kimi-moonshot alibaba ollama zhipu-ai stepfun parallel-thinking model-releases moe attention-mechanisms multimodal-reasoning model-performance context-windows open-source-models model-leaks creative-ai coding reasoning model-optimization demishassabis philschmid scaling01 teortaxestex teknium1 lmarena_ai andrewyng
OpenAI is rumored to soon launch new GPT-OSS and GPT-5 models amid drama with Anthropic revoking access to Claude. Google DeepMind quietly launched Gemini 2.5 Deep Think, a model optimized for parallel thinking that achieved gold-medal level at the IMO and excels in reasoning, coding, and creative tasks. Leaks suggest OpenAI is developing a 120B MoE and a 20B model with advanced attention mechanisms. Chinese AI companies like Kimi Moonshot, Alibaba, and ZHIpu AI are releasing faster and more capable open models such as kimi-k2-turbo-preview, Qwen3-Coder-Flash, and GLM-4.5, signaling strong momentum and potential to surpass the U.S. in AI development. "The final checkpoint was selected just 5 hours before the IMO problems were released," highlighting rapid development cycles.
GLM-4.5: Deeper, Headier, & better than Kimi/Qwen/DeepSeek (SOTA China LLM?)
glm-4.5-355b-a32b glm-4.5-air-106b-a12b qwen3-coder claude-4-opus grok-4 o3 gpt-4.1 gpt-5 kimi-k2 claude-sonnet-4 z-ai alibaba huggingface openai reinforcement-learning token-efficiency model-optimization open-source-models agentic-ai coding model-training lupantech teortaxestex mervenoyann _lewtun scaling01 cline
Z.ai (Zhipu AI) released the GLM-4.5-355B-A32B and GLM-4.5-Air-106B-A12B open weights models, claiming state-of-the-art performance competitive with Claude 4 Opus, Grok 4, and OpenAI's o3. These models emphasize token efficiency and efficient reinforcement learning training validated by the Muon optimizer. Alibaba Qwen introduced Group Sequence Policy Optimization (GSPO), a new reinforcement learning algorithm powering the Qwen3 model suite, integrated into Hugging Face's TRL library. Speculation surrounds mystery models "summit" and "zenith" as potential GPT-5 variants based on GPT-4.1 architecture. Qwen3-Coder shows strong coding benchmark results, rivaling Claude Sonnet 4 and Kimi K2. The rise of powerful Chinese open-source models like GLM-4.5, Wan-2.2, and Qwen3 Coder contrasts with a slowdown from Western labs such as OpenAI.
not much happened today
gpt-5 gpt4-0314 qwen3-235b-thinking runway-aleph imagen-4-ultra smollm3 grok-4 openai alibaba runway hugging-face google anthropic pytorch lmarena reinforcement-learning reasoning video-generation image-generation model-optimization open-source model-performance inference-speed integration stability sama clementdelangue xikun_zhang_ teknnium1 chujiezheng
OpenAI has fully rolled out its ChatGPT agent to all Plus, Pro, and Team users and is building hype for the upcoming GPT-5, which reportedly outperforms Grok-4 and can build a cookie clicker game in two minutes. Alibaba's Qwen team released the open-source reasoning model Qwen3-235B-Thinking, achieving an 89% win rate over gpt4-0314 using a new RL algorithm called Group Sequence Policy Optimization (GSPO). Runway introduced Runway Aleph, a state-of-the-art in-context video model for editing and generating video content. Hugging Face highlights the growing momentum of open-source AI, especially from Chinese teams. Other updates include Kling's upgrades for image-to-video generation and Google's Imagen 4 Ultra being recognized as a top text-to-image model. Anthropic integrated Claude with Canva for branded visual designs but faces stability issues. The PyTorch team released optimized checkpoints for SmolLM3 to speed up inference.
not much happened today
kimi-k2 qwen3-235b-a22b qwen3-coder-480b-a35b gemini-2.5-flash-lite mistral-7b deepseek-v3 moonshot-ai alibaba google google-deepmind openai hugging-face vllm-project mixture-of-experts agentic-ai model-optimization model-training benchmarking code-generation long-context multimodality math reinforcement-learning model-architecture model-performance open-source alignment demishassabis rasbt alexwei_ yitayml
Moonshot AI released the Kimi K2, a 1-trillion parameter ultra-sparse Mixture-of-Experts (MoE) model with the MuonClip optimizer and a large-scale agentic data pipeline using over 20,000 tools. Shortly after, Alibaba updated its Qwen3 model with the Qwen3-235B-A22B variant, which outperforms Kimi K2 and other top models on benchmarks like GPQA and AIME despite being 4.25x smaller. Alibaba also released Qwen3-Coder-480B-A35B, a MoE model specialized for coding with a 1 million token context window. Google DeepMind launched Gemini 2.5 Flash-Lite, a faster and more cost-efficient model outperforming previous versions in coding, math, and multimodal tasks. The MoE architecture is becoming mainstream, with models like Mistral, DeepSeek, and Kimi K2 leading the trend. In mathematics, an advanced Gemini model achieved a gold medal level score at the International Mathematical Olympiad (IMO), marking a first for AI. An OpenAI researcher noted their IMO model "knew" when it did not have a correct solution, highlighting advances in model reasoning and self-awareness.
Kimi K2 - SOTA Open MoE proves that Muon can scale to 15T tokens/1T params
kimi-k2 kimi-k2-1t deepseek-v3 grok-4 devstral-2507 gpt-4.1 sonnet-4 moonshot-ai alibaba tencent deepseek x-ai mistral-ai weights-biases hugging-face mixture-of-experts model-training model-optimization optimizer benchmarking long-context model-performance open-weights model-release yuchenj_uw andrew_n_carr scaling01 novita_labs teknium1 aravsrinivas mparakhin simonw
Moonshot AI has released Kimi K2, a 1 trillion parameter Mixture-of-Experts model trained on 15.5 trillion tokens using the new MuonClip optimizer, achieving state-of-the-art results on benchmarks like SWE-Bench Verified (65.8%) and TAU2 (58.4%). This model is competitive with GPT-4.1 and Sonnet 4 on non-thinking tasks and is available under an MIT license. Meanwhile, xAI announced Grok-4, noted for its "LEAST censored frontier model" status and strong long-context performance but criticized for rushed post-training. Mistral AI updated its Devstral 2507 models with improved performance and cost efficiency. The community is excited about the potential of the MuonClip optimizer, which may surpass the long-standing AdamW optimizer in machine learning.
not much happened today
o3-mini o1-mini llama hunyuan-a13b ernie-4.5 ernie-4.5-21b-a3b qwen3-30b-a3b gemini-2.5-pro meta-ai-fair openai tencent microsoft baidu gemini superintelligence ai-talent job-market open-source-models multimodality mixture-of-experts quantization fp8-training model-benchmarking model-performance model-releases api model-optimization alexandr_wang shengjia_zhao jhyuxm ren_hongyu shuchaobi saranormous teortaxesTex mckbrando yuchenj_uw francoisfleuret quanquangu reach_vb philschmid
Meta has poached top AI talent from OpenAI, including Alexandr Wang joining as Chief AI Officer to work towards superintelligence, signaling a strong push for the next Llama model. The AI job market shows polarization with high demand and compensation for top-tier talent, while credentials like strong GitHub projects gain importance. The WizardLM team moved from Microsoft to Tencent to develop open-source models like Hunyuan-A13B, highlighting shifts in China's AI industry. Rumors suggest OpenAI will release a new open-source model in July, potentially surpassing existing ChatGPT models. Baidu open-sourced multiple variants of its ERNIE 4.5 model series, featuring advanced techniques like 2-bit quantization, MoE router orthogonalization loss, and FP8 training, with models ranging from 0.3B to 424B parameters. Gemini 2.5 Pro returned to the free tier of the Gemini API, enabling developers to explore its features.
OpenAI releases Deep Research API (o3/o4-mini)
o3-deep-research o4-mini-deep-research gemma-3n flux-1-kontext-dev gpt-4o alphagenome openai google black-forest-labs deepmind sakana-ai higgsfield-ai huggingface ollama multimodality model-releases agentic-ai reinforcement-learning instruction-following model-architecture model-optimization image-generation biological-ai multi-agent-systems model-integration demishassabis hardmaru osanseviero clementdelangue
OpenAI has launched the Deep Research API featuring powerful models o3-deep-research and o4-mini-deep-research with native support for MCP, Search, and Code Interpreter, enabling advanced agent capabilities including multi-agent setups. Google released Gemma 3n, a multimodal model optimized for edge devices with only 3GB RAM, achieving a top score of 1300 on LMSys Arena, featuring the new MatFormer architecture and broad ecosystem integration. Black Forest Labs introduced FLUX.1 Kontext [dev], a 12B parameter rectified flow transformer for instruction-based image editing, comparable to GPT-4o. DeepMind unveiled AlphaGenome, an AI model capable of reading 1 million DNA bases for gene function prediction, marking a breakthrough in AI biology. Sakana AI presented Reinforcement-Learned Teachers (RLTs) to enhance LLM reasoning, achieving 86.1% on MiniF2F with efficient compute. Higgsfield AI released Higgsfield Soul, a high-aesthetic photo model with 50+ presets for fashion-grade realism. Additionally, Google launched the Gemini CLI, an open-source AI agent for terminal use with free Gemini 2.5 Pro requests.
not much happened today
dots-llm1 qwen3-235b xiaohongshu rednote-hilab deepseek huggingface mixture-of-experts open-source model-benchmarking fine-tuning inference context-windows training-data model-architecture model-performance model-optimization
China's Xiaohongshu (Rednote) released dots.llm1, a 142B parameter open-source Mixture-of-Experts (MoE) language model with 14B active parameters and a 32K context window, pretrained on 11.2 trillion high-quality, non-synthetic tokens. The model supports efficient inference frameworks like Docker, HuggingFace, and vLLM, and provides intermediate checkpoints every 1 trillion tokens, enabling flexible fine-tuning. Benchmarking claims it slightly surpasses Qwen3 235B on MMLU, though some concerns exist about benchmark selection and synthetic data verification. The release is notable for its truly open-source licensing and no synthetic data usage, sparking community optimism for support in frameworks such as llama.cpp and mlx.
Mary Meeker is so back: BOND Capital AI Trends report
qwen-3-8b anthropic hugging-face deepseek attention-mechanisms inference arithmetic-intensity transformers model-optimization interpretability model-quantization training tri_dao fleetwood___ teortaxestex awnihannun lateinteraction neelnanda5 eliebakouch _akhaliq
Mary Meeker returns with a comprehensive 340-slide report on the state of AI, highlighting accelerating tech cycles, compute growth, and comparisons of ChatGPT to early Google and other iconic tech products. The report also covers enterprise traction and valuation of major AI companies. On Twitter, @tri_dao discusses an "ideal" inference architecture featuring attention variants like GTA, GLA, and DeepSeek MLA with high arithmetic intensity (~256), improving efficiency and model quality. Other highlights include the release of 4-bit DWQ of DSR1 Qwen3 8B on Hugging Face, AnthropicAI's open-source interpretability tools for LLMs, and discussions on transformer training and abstractions by various researchers.
not much happened today
deepseek-r1-0528 pali-gemma-2 gemma-3 shieldgemma-2 txgemma gemma-3-qat gemma-3n-preview medgemma dolphingemma signgemma claude-4 opus-4 claude-sonnet-4 codestral-embed bagel qwen nemotron-cortexa gemini-2.5-pro deepseek-ai huggingface gemma claude bytedance qwen nemotron sakana-ai-labs benchmarking model-releases multimodality code-generation model-performance long-context reinforcement-learning model-optimization open-source yuchenj_uw _akhaliq clementdelangue osanseviero alexalbert__ guillaumelample theturingpost lmarena_ai epochairesearch scaling01 nrehiew_ ctnzr
DeepSeek R1 v2 model released with availability on Hugging Face and inference partners. The Gemma model family continues prolific development including PaliGemma 2, Gemma 3, and others. Claude 4 and its variants like Opus 4 and Claude Sonnet 4 show top benchmark performance, including new SOTA on ARC-AGI-2 and WebDev Arena. Codestral Embed introduces a 3072-dimensional code embedder. BAGEL, an open-source multimodal model by ByteDance, supports reading, reasoning, drawing, and editing with long mixed contexts. Benchmarking highlights include Nemotron-CORTEXA topping SWEBench and Gemini 2.5 Pro performing on VideoGameBench. Discussions on random rewards effectiveness focus on Qwen models. "Opus 4 NEW SOTA ON ARC-AGI-2. It's happening - I was right" and "Claude 4 launch has dev moving at a different pace" reflect excitement in the community.
Google I/O: new Gemini native voice, Flash, DeepThink, AI Mode (DeepSearch+Mariner+Astra)
gemini-2.5-pro gemini-2.5 google google-deepmind ai-assistants reasoning generative-ai developer-tools ai-integration model-optimization ai-application model-updates ai-deployment model-performance demishassabis philschmid jack_w_rae
Google I/O 2024 showcased significant advancements with Gemini 2.5 Pro and Deep Think reasoning mode from google-deepmind, emphasizing AI-driven transformations and developer opportunities. GeminiApp aims to become a universal AI assistant on the path to AGI, with new features like AI Mode in Google Search expanding generative AI access. The event included multiple keynotes and updates on over a dozen models and 20+ AI products, highlighting Google's leadership in AI innovation. Influential voices like demishassabis and philschmid provided insights and recaps, while the launch of Jules as a competitor to Codex/Devin was noted.
not much happened today
hunyuan-turbos qwen3-235b-a22b o3 gpt-4.1-nano grok-3 gemini-2.5-pro seed1.5-vl kling-2.0 tencent openai bytedance meta-ai-fair nvidia deepseek benchmarking model-performance moe reasoning vision video-understanding vision-language multimodality model-evaluation model-optimization lmarena_ai artificialanlys gdb _jasonwei iScienceLuvr _akhaliq _philschmid teortaxesTex mervenoyann reach_vb
Tencent's Hunyuan-Turbos has risen to #8 on the LMArena leaderboard, showing strong performance across major categories and significant improvement since February. The Qwen3 model family, especially the Qwen3 235B-A22B (Reasoning) model, is noted for its intelligence and efficient parameter usage. OpenAI introduced HealthBench, a new health evaluation benchmark developed with input from over 250 physicians, where models like o3, GPT-4.1 nano, and Grok 3 showed strong results. ByteDance released Seed1.5-VL, a vision-language model with a 532M-parameter vision encoder and a 20B active parameter MoE LLM, achieving state-of-the-art results on 38 public benchmarks. In vision-language, Kling 2.0 leads image-to-video generation, and Gemini 2.5 Pro excels in video understanding with advanced multimodal capabilities. Meta's Vision-Language-Action framework and updates on VLMs for 2025 were also highlighted.
Prime Intellect's INTELLECT-2 and PRIME-RL advance distributed reinforcement learning
intellect-2 dreamo qwen gemini-2.5-pro dynamic-byte-latent-transformer gen-4-references mistral-medium-3 le-chat-enterprise primeintellect bytedance qwen gemma meta-ai-fair runwayml mistral-ai google distributed-training reinforcement-learning gpu-clusters model-optimization quantization multimodality agentic-ai video-understanding fine-tuning _akhaliq reach_vb osanseviero aiatmeta c_valenzuelab lmarena_ai adcock_brett
Prime Intellect released INTELLECT-2, a decentralized GPU training and RL framework with a vision for distributed AI training overcoming colocation limits. ByteDance launched DreamO, a unified image customization model on Hugging Face. Qwen released models optimized for GPTQ, GGUF, and AWQ quantization. Gemma surpassed 150 million downloads on Hugging Face. Meta released weights for the Dynamic Byte Latent Transformer and the Collaborative Reasoner framework to improve language model efficiency and reasoning. RunwayML introduced Gen-4 References, a near-realtime model requiring no fine-tuning. Mistral AI released Mistral Medium 3, a strong multimodal model, and Le Chat Enterprise, an agentic AI assistant for business. Google updated Gemini 2.5 Pro Preview with video understanding and UI improvements. "Airbnb for spare GPUs from all over the world" highlights the ongoing challenges and potential of distributed GPU training.
not much happened today
open-code-reasoning-32b open-code-reasoning-14b open-code-reasoning-7b mistral-medium-3 llama-4-maverick gemini-2.5-pro gemini-2.5-flash claude-3.7-sonnet absolute-zero-reasoner x-reasoner fastvlm parakeet-asr openai nvidia mistral-ai google apple huggingface reinforcement-learning fine-tuning code-generation reasoning vision on-device-ai model-performance dataset-release model-optimization reach_vb artificialanlys scaling01 iscienceluvr arankomatsuzaki awnihannun risingsayak
OpenAI launched both Reinforcement Finetuning and Deep Research on GitHub repos, drawing comparisons to Cognition's DeepWiki. Nvidia open-sourced Open Code Reasoning models (32B, 14B, 7B) with Apache 2.0 license, showing 30% better token efficiency and compatibility with llama.cpp, vLLM, transformers, and TGI. Independent evaluations highlight Mistral Medium 3 rivaling Llama 4 Maverick, Gemini 2.0 Flash, and Claude 3.7 Sonnet in coding and math reasoning, priced significantly lower but no longer open-source. Google's Gemini 2.5 Pro is noted as their most intelligent model with improved coding from simple prompts, while Gemini 2.5 Flash incurs a 150x cost increase over Gemini 2.0 Flash due to higher token usage and cost. The Absolute Zero Reasoner (AZR) achieves SOTA performance in coding and math reasoning via reinforced self-play without external data. Vision-language model X-REASONER is post-trained on general-domain text for reasoning. Apple ML research released FastVLM with on-device iPhone demo. HiDream LoRA trainer supports QLoRA fine-tuning under memory constraints. Nvidia's Parakeet ASR model tops Hugging Face ASR leaderboard with MLX implementation. New datasets SwallowCode and SwallowMath boost LLM performance in math and code. Overall, a quiet day with significant model releases and performance insights.
ChatGPT responds to GlazeGate + LMArena responds to Cohere
qwen3-235b-a22b qwen3 qwen3-moe llama-4 openai cohere lm-arena deepmind x-ai meta-ai-fair alibaba vllm llamaindex model-releases model-benchmarking performance-evaluation open-source multilinguality model-integration fine-tuning model-optimization joannejang arankomatsuzaki karpathy sarahookr reach_vb
OpenAI faced backlash after a controversial ChatGPT update, leading to an official retraction admitting they "focused too much on short-term feedback." Researchers from Cohere published a paper criticizing LMArena for unfair practices favoring incumbents like OpenAI, DeepMind, X.ai, and Meta AI Fair. The Qwen3 family by Alibaba was released, featuring models up to 235B MoE, supporting 119 languages and trained on 36 trillion tokens, with integration into vLLM and support in tools like llama.cpp. Meta announced the second round of Llama Impact Grants to promote open-source AI innovation. Discussions on AI Twitter highlighted concerns about leaderboard overfitting and fairness in model benchmarking, with notable commentary from karpathy and others.
LlamaCon: Meta AI gets into the Llama API platform business
llama-4 qwen3 qwen3-235b-a22b qwen3-30b-a3b qwen3-4b qwen2-5-72b-instruct o3-mini meta-ai-fair cerebras groq alibaba vllm ollama llamaindex hugging-face llama-cpp model-release fine-tuning reinforcement-learning moe multilingual-models model-optimization model-deployment coding benchmarking apache-license reach_vb huybery teortaxestex awnihannun thezachmueller
Meta celebrated progress in the Llama ecosystem at LlamaCon, launching an AI Developer platform with finetuning and fast inference powered by Cerebras and Groq hardware, though it remains waitlisted. Meanwhile, Alibaba released the Qwen3 family of large language models, including two MoE models and six dense models ranging from 0.6B to 235B parameters, with the flagship Qwen3-235B-A22B achieving competitive benchmark results and supporting 119 languages and dialects. The Qwen3 models are optimized for coding and agentic capabilities, are Apache 2.0 licensed, and have broad deployment support including local usage with tools like vLLM, Ollama, and llama.cpp. Community feedback highlights Qwen3's scalable performance and superiority over models like OpenAI's o3-mini.
not much happened today
nemotron-h nvidia-eagle-2.5 gpt-4o qwen2.5-vl-72b gemini-2.5-flash gemini-2.0-pro gemini-exp-1206 gemma-3 qwen2.5-32b deepseek-r1-zero-32b uni3c seedream-3.0 adobe-dragon kimina-prover qwen2.5-72b bitnet-b1.58-2b4t nvidia deepseek hugging-face alibaba bytedance adobe transformers model-optimization multimodality long-context reinforcement-learning torch-compile image-generation diffusion-models distributional-rewards model-efficiency model-training native-quantization sampling-techniques philschmid arankomatsuzaki osanseviero iScienceLuvr akhaliq
Nemotron-H model family introduces hybrid Mamba-Transformer models with up to 3x faster inference and variants including 8B, 56B, and a compressed 47B model. Nvidia Eagle 2.5 is a frontier VLM for long-context multimodal learning, matching GPT-4o and Qwen2.5-VL-72B on long-video understanding. Gemini 2.5 Flash shows improved dynamic thinking and cost-performance, outperforming previous Gemini versions. Gemma 3 now supports torch.compile for about 60% faster inference on consumer GPUs. SRPO using Qwen2.5-32B surpasses DeepSeek-R1-Zero-32B on benchmarks with reinforcement learning only. Alibaba's Uni3C unifies 3D-enhanced camera and human motion controls for video generation. Seedream 3.0 by ByteDance is a bilingual image generation model with high-resolution outputs up to 2K. Adobe DRAGON optimizes diffusion generative models with distributional rewards. Kimina-Prover Preview is an LLM trained with reinforcement learning from Qwen2.5-72B, achieving 80.7% pass@8192 on miniF2F. BitNet b1.58 2B4T is a native 1-bit LLM with 2B parameters trained on 4 trillion tokens, matching full-precision LLM performance with better efficiency. Antidistillation sampling counters unwanted model distillation by modifying reasoning traces from frontier models.
Cohere's Command A claims #3 open model spot (after DeepSeek and Gemma)
command-a mistral-ai-small-3.1 smoldocling qwen-2.5-vl cohere mistral-ai hugging-face context-windows multilinguality multimodality fine-tuning benchmarking ocr model-performance model-releases model-optimization aidangomez sophiamyang mervenoyann aidan_mclau reach_vb lateinteraction
Cohere's Command A model has solidified its position on the LMArena leaderboard, featuring an open-weight 111B parameter model with an unusually long 256K context window and competitive pricing. Mistral AI released the lightweight, multilingual, and multimodal Mistral AI Small 3.1 model, optimized for single RTX 4090 or Mac 32GB RAM setups, with strong performance on instruct and multimodal benchmarks. The new OCR model SmolDocling offers fast document reading with low VRAM usage, outperforming larger models like Qwen2.5VL. Discussions highlight the importance of system-level improvements over raw LLM advancements, and MCBench is recommended as a superior AI benchmark for evaluating model capabilities across code, aesthetics, and awareness.
DeepSeek's Open Source Stack
qwen-qwq-32b start character-3 gemini gemini-2.0 mercury-coder gpt-4.5 jamba-mini-1.6 gemini-2.0-flash gpt-4o-mini mistral-small-3 mistral-ocr deepseek pyspur hugging-face togethercompute hedra-labs google-deepmind deeplearningai openai ai21-labs mistral-ai fine-tuning benchmarking multimodality code-generation diffusion-models model-performance model-optimization ocr embedding-models context-windows runtime-limits _akhaliq lmarena_ai reach_vb danielhanchen _philschmid aidan_mclau vikhyatk jerryjliu0
DeepSeek's Open Source Week was summarized by PySpur, highlighting multiple interesting releases. The Qwen QwQ-32B model was fine-tuned into START, excelling in PhD-level science QA and math benchmarks. Character-3, an omnimodal AI video generation model by Hedra Labs and Together AI, enables realistic animated content creation. Google DeepMind introduced the Gemini embedding model with an 8k context window, ranking #1 on MMTEB, alongside the Gemini 2.0 Code Executor supporting Python libraries and auto-fix features. Inception Labs' Mercury Coder is a diffusion-based code generation model offering faster token processing. OpenAI released GPT-4.5, their largest model yet but with less reasoning ability than some competitors. AI21 Labs launched Jamba Mini 1.6, noted for superior output speed compared to Gemini 2.0 Flash, GPT-4o mini, and Mistral Small 3. A new dataset of 1.9M scanned pages was released for OCR benchmarking, with Mistral OCR showing competitive but not top-tier document parsing performance compared to LLM/LVM-powered methods. "Cracked engineers are all you need."
not much happened today
chatgpt-4o deepseek-r1 o3 o3-mini gemini-2-flash qwen-2.5 qwen-0.5b hugging-face openai perplexity-ai deepseek-ai gemini qwen metr_evals reasoning benchmarking model-performance prompt-engineering model-optimization model-deployment small-language-models mobile-ai ai-agents speed-optimization _akhaliq aravsrinivas lmarena_ai omarsar0 risingsayak
Smolagents library by Huggingface continues trending. ChatGPT-4o latest version
chatgpt-40-latest-20250129 released. DeepSeek R1 671B sets speed record at 198 t/s, fastest reasoning model, recommended with specific prompt settings. Perplexity Deep Research outperforms models like Gemini Thinking, o3-mini, and DeepSeek-R1 on Humanity's Last Exam benchmark with 21.1% score and 93.9% accuracy on SimpleQA. ChatGPT-4o ranks #1 on Arena leaderboard in multiple categories except math. OpenAI's o3 model powers Deep Research tool for ChatGPT Pro users. Gemini 2 Flash and Qwen 2.5 models support LLMGrading verifier. Qwen 2.5 models added to PocketPal app. MLX shows small LLMs like Qwen 0.5B generate tokens at high speed on M4 Max and iPhone 16 Pro. Gemini Flash 2.0 leads new AI agent leaderboard. DeepSeek R1 is most liked on Hugging Face with over 10 million downloads. Mistral Small 3 24B and Tulu 3 405B
mistral-small-3 tulu-3-405b llama-3 tiny-swallow-1.5b qwen-2.5-max deepseek-v3 claude-3.5-sonnet gemini-1.5-pro gpt4o-mini llama-3-3-70b mistral-ai ai2 sakana-ai alibaba_qwen deepseek ollama llamaindex reinforcement-learning model-fine-tuning local-inference model-performance model-optimization on-device-ai instruction-following api training-data natural-language-processing clementdelangue dchaplot reach_vb
Mistral AI released Mistral Small 3, a 24B parameter model optimized for local inference with low latency and 81% accuracy on MMLU, competing with Llama 3.3 70B, Qwen-2.5 32B, and GPT4o-mini. AI2 released Tülu 3 405B, a large finetuned model of Llama 3 using Reinforcement Learning from Verifiable Rewards (RVLR), competitive with DeepSeek v3. Sakana AI launched TinySwallow-1.5B, a Japanese language model using TAID for on-device use. Alibaba_Qwen released Qwen 2.5 Max, trained on 20 trillion tokens, with performance comparable to DeepSeek V3, Claude 3.5 Sonnet, and Gemini 1.5 Pro, and updated API pricing. These releases highlight advances in open models, efficient inference, and reinforcement learning techniques.
DeepSeek R1: o1-level open weights model and a simple recipe for upgrading 1.5B models to Sonnet/4o level
deepseek-r1 deepseek-v3 qwen-2.5 llama-3.1 llama-3.3-70b deepseek ollama qwen llama reinforcement-learning fine-tuning model-distillation model-optimization reasoning reward-models multi-response-sampling model-training
DeepSeek released DeepSeek R1, a significant upgrade over DeepSeek V3 from just three weeks prior, featuring 8 models including full-size 671B MoE models and multiple distillations from Qwen 2.5 and Llama 3.1/3.3. The models are MIT licensed, allowing finetuning and distillation. Pricing is notably cheaper than o1 by 27x-50x. The training process used GRPO (reward for correctness and style outcomes) without relying on PRM, MCTS, or reward models, focusing on reasoning improvements through reinforcement learning. Distilled models can run on Ollama and show strong capabilities like writing Manim code. The release emphasizes advances in reinforcement-learning, fine-tuning, and model-distillation with a novel RL framework from DeepSeekMath.
not much happened today
helium-1 qwen-2.5 phi-4 sky-t1-32b-preview o1 codestral-25.01 phi-3 mistral llama-3 gpt-3.5 llama-3 gpt-3.5 llmquoter kyutai-labs lmstudio mistralai llamaindex huggingface langchainai hyperbolic-labs replit fchollet philschmid multilinguality token-level-distillation context-windows model-performance open-source reasoning coding retrieval-augmented-generation hybrid-retrieval multiagent-systems video large-video-language-models dynamic-ui voice-interaction gpu-rentals model-optimization semantic-deduplication model-inference reach_vb awnihannun lior_on_ai sophiamyang omarsar0 skirano yuchenj_uw fchollet philschmid
Helium-1 Preview by kyutai_labs is a 2B-parameter multilingual base LLM outperforming Qwen 2.5, trained on 2.5T tokens with a 4096 context size using token-level distillation from a 7B model. Phi-4 (4-bit) was released in lmstudio on an M4 max, noted for speed and performance. Sky-T1-32B-Preview is a $450 open-source reasoning model matching o1's performance with strong benchmark scores. Codestral 25.01 by mistralai is a new SOTA coding model supporting 80+ programming languages and offering 2x speed.
Innovations include AutoRAG for optimizing retrieval-augmented generation pipelines, Agentic RAG for autonomous query reformulation and critique, Multiagent Finetuning using societies of models like Phi-3, Mistral, LLaMA-3, and GPT-3.5 for reasoning improvements, and VideoRAG incorporating video content into RAG with LVLMs.
Applications include a dynamic UI AI chat app by skirano on Replit, LangChain tools like DocTalk for voice PDF conversations, AI travel agent tutorials, and news summarization agents. Hyperbolic Labs offers competitive GPU rentals including H100, A100, and RTX 4090. LLMQuoter enhances RAG accuracy by identifying key quotes.
Infrastructure updates include MLX export for LLM inference from Python to C++ by fchollet and SemHash semantic text deduplication by philschmid.
not much happened today
phi-4 reinforce++ arc-agi-2 ai21-labs ollama langchain togethercompute groq reinforcement-learning ppo model-optimization memory-efficiency python-packages vision text-extraction frontend-code-generation workflow-automation coding-agents compute-cost-reduction ethical-ai agi-benchmarks scam-alerts sebastien-bubeck fchollet tom-doerr arohan_ bindureddy hwchase17 jonathanross321 clementdelangue vikhyatk
Sebastien Bubeck introduced REINFORCE++, enhancing classical REINFORCE with PPO-inspired techniques for 30% faster training. AI21 Labs released Phi-4 under the MIT License, accessible via Ollama. François Chollet announced plans for ARC-AGI-2 and a next-generation AGI benchmark. LangChain launched 10 new integration packages to boost LLM application development. Tom Doerr introduced Ollama-OCR, a Python package for text extraction using vision language models. Arohan optimized Shampoo for memory efficiency, reducing usage from 20 to 6 bytes per parameter. Bindu Reddy showcased CodeLLM's v1 for frontend code generation and highlighted LlamaIndex Workflows for academic summarization and slide generation. Hwchase17 collaborated with Together Compute to enhance WebDev Arena with complex coding agents for LLM coding evaluations. Jonathan Ross detailed Groq's mission to reduce compute costs by 1000x amid rising generative AI spending. Clement Delangue warned about scam alerts involving false claims of association with AI21. Vikhyat K raised concerns about the ethical implications and trade-offs of AGI. Memes and humor included creative AI prompts and critiques of LLM behaviors.
DeepSeek v3: 671B finegrained MoE trained for $5.5m USD of compute on 15T tokens
deepseek-v3 gpt-4o claude-3.5-sonnet llama-3 deepseek-ai hugging-face openai anthropic mixture-of-experts model-training model-optimization reinforcement-learning chain-of-thought multi-token-prediction synthetic-data model-distillation fine-tuning attention-mechanisms gpu-optimization nrehiew_ denny_zhou
DeepSeek-V3 has launched with 671B MoE parameters and trained on 14.8T tokens, outperforming GPT-4o and Claude-3.5-sonnet in benchmarks. It was trained with only 2.788M H800 GPU hours, significantly less than Llama-3's 30.8M GPU-hours, showcasing major compute efficiency and cost reduction. The model is open-source and deployed via Hugging Face with API support. Innovations include native FP8 mixed precision training, Multi-Head Latent Attention scaling, distillation from synthetic reasoning data, pruning and healing for MoEs with up to 256 experts, and a new multi-token prediction objective enabling lookahead token planning. Research highlights also cover the OREO method and Natural Language Reinforcement Learning (NLRL) for multi-step reasoning and agent control.
Meta BLT: Tokenizer-free, Byte-level LLM
byte-latent-transformer llama-3 phi-4 gpt-4o command-r7b meta-ai-fair llamaindex microsoft deepseek-ai openai cohere anthropic tokenization transformer-architecture model-efficiency benchmarking multimodality vision reinforcement-learning model-scaling jailbreaking model-optimization
Meta AI introduces the Byte Latent Transformer (BLT), a tokenizer-free architecture that dynamically forms byte patches for efficient compute allocation, outperforming Llama 3 on benchmarks including the CUTE benchmark. The model was trained on approximately 1 trillion tokens and features a three-block transformer design with local and global components. This approach challenges traditional tokenization and may enable new multimodal capabilities such as direct file interaction without retrieval-augmented generation. Additionally, Microsoft announced the Phi-4 14B parameter model achieving state-of-the-art results on STEM and reasoning benchmarks, surpassing GPT-4o. DeepSeek AI launched new vision-language models based on their MoE architecture with sizes ranging from 1.0B to 27B parameters. OpenAI released a new Projects feature for ChatGPT, and Cohere introduced their smallest and fastest Command R7B model. Anthropic published research on "Best-of-N Jailbreaking" vulnerabilities across text, vision, and audio models. Industry discussion highlights a trend of decreasing frontier LLM sizes, with GPT-4 at approximately 1.8 trillion parameters compared to newer models.
not much happened today
o1-full sora gpt-4.5 gpt-4 claude-3.5-sonnet llama-3-1-nemotron-51b llama-3-1 llama-3 nemotron-51b openai google-deepmind anthropic nvidia huggingface vision model-performance neural-architecture-search model-optimization multimodality model-release model-training reinforcement-learning image-generation lucas-beyer alexander-kolesnikov xiaohua-zhai aidan_mclau giffmana joannejang sama
OpenAI announced their "12 Days of OpenAI" event with daily livestreams and potential releases including the O1 full model, Sora video model, and GPT-4.5. Google DeepMind released the GenCast weather model capable of 15-day forecasts in 8 minutes using TPU chips, and launched Genie 2, a model generating playable 3D worlds from single images. Leading vision researchers Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai moved from DeepMind to OpenAI, which is opening a Zürich office. Criticism arose over OpenAI's strategy and model quality compared to Anthropic and Claude 3.5 Sonnet. On Reddit, a modified llama.cpp supports Nvidia's Llama-3_1-Nemotron-51B, matching performance of larger 70B models via NAS optimization.
Olympus has dropped (aka, Amazon Nova Micro|Lite|Pro|Premier|Canvas|Reel)
amazon-nova claude-3 llama-3-70b gemini-1.5-flash gpt-4o amazon anthropic google-deepmind sakana-ai-labs multimodality benchmarking model-merging model-performance model-architecture model-optimization population-based-learning philschmid bindureddy
Amazon announced the Amazon Nova family of multimodal foundation models at AWS Re:Invent, available immediately with no waitlist in configurations like Micro, Lite, Pro, Canvas, and Reel, with Premier and speech-to-speech coming next year. These models offer 2-4x faster token speeds and are 25%-400% cheaper than competitors like Anthropic Claude models, positioning Nova as a serious contender in AI engineering. Pricing undercuts models such as Google DeepMind Gemini Flash 8B, and some Nova models extend context length up to 300k tokens. However, benchmarking controversy exists as some evaluations show Nova scoring below Llama-3 70B in LiveBench AI metrics. Separately, CycleQD was introduced by Sakana AI Labs, using evolutionary computation for population-based model merging to develop niche LLM agents.
Vision Everywhere: Apple AIMv2 and Jina CLIP v2
aimv2-3b jina-clip-v2 tulu-3 llama-3-1 claude-3-5 llama-3-1-70b apple jina allen_ai autoregressive-objectives vision multilinguality multimodality image-generation model-training model-optimization reinforcement-learning fine-tuning model-benchmarking
Apple released AIMv2, a novel vision encoder pre-trained with autoregressive objectives that achieves 89.5% accuracy on ImageNet and integrates joint visual and textual objectives. Jina launched Jina CLIP v2, a multimodal embedding model supporting 89 languages and high-resolution images with efficient Matryoshka embeddings reducing dimensions by 94% with minimal accuracy loss. Allen AI introduced Tülu 3 models based on Llama 3.1 with 8B and 70B parameters, offering 2.5x faster inference and alignment via SFT, DPO, and RLVR methods, competing with Claude 3.5 and Llama 3.1 70B. These developments highlight advances in autoregressive training, vision encoders, and multilingual multimodal embeddings.
Tencent's Hunyuan-Large claims to beat DeepSeek-V2 and Llama3-405B with LESS Data
claude-3.5-haiku llama-3-1 llama-3-2 mlx-lm tencent anthropic meta-ai-fair togethercompute llamaindex mixture-of-experts synthetic-data model-scaling model-architecture model-optimization kv-cache-quantization react fine-tuning scaling-laws model-efficiency model-deployment multimodality
Tencent released a notable >300B parameter MoE model pretrained on 7T tokens, including 1.5T synthetic data generated via Evol-Instruct. The model introduces novel techniques like "recycle routing" and expert-specific learning rates, alongside a compute-efficient scaling law for MoE active parameters. However, its custom license restricts use in the EU and by companies with over 100M MAU, and it avoids China-sensitive queries. Meanwhile, Anthropic launched Claude 3.5 Haiku, now available on multiple platforms, praised for intelligence and speed but criticized for a 10x price increase. Meta opened Llama AI to the U.S. defense sector, and a Llama Impact Hackathon offers a $15K prize for projects using Llama 3.1 & 3.2 Vision. LlamaIndex released a React chat UI component with Tailwind CSS and LLM backend integrations. The MLX LM model advances text generation speed and efficiency with KV cache quantization.
not much happened today
smollm2 llama-3-2 stable-diffusion-3.5 claude-3.5-sonnet gemini openai anthropic google meta-ai-fair suno-ai perplexity-ai on-device-ai model-performance robotics multimodality ai-regulation model-releases natural-language-processing prompt-engineering agentic-ai ai-application model-optimization sam-altman akhaliq arav-srinivas labenz loubnabenallal1 alexalbert fchollet stasbekman svpino rohanpaul_ai hamelhusain
ChatGPT Search was launched by Sam Altman, who called it his favorite feature since ChatGPT's original launch, doubling his usage. Comparisons were made between ChatGPT Search and Perplexity with improvements noted in Perplexity's web navigation. Google introduced a "Grounding" feature in the Gemini API & AI Studio enabling Gemini models to access real-time web information. Despite Gemini's leaderboard performance, developer adoption lags behind OpenAI and Anthropic. SmolLM2, a new small, powerful on-device language model, outperforms Meta's Llama 3.2 1B. A Claude desktop app was released for Mac and Windows. Meta AI announced robotics advancements including Meta Sparsh, Meta Digit 360, and Meta Digit Plexus. Stable Diffusion 3.5 Medium, a 2B parameter model with a permissive license, was released. Insights on AGI development suggest initial inferiority but rapid improvement. Anthropic advocates for early targeted AI regulation. Discussions on ML specialization predict training will concentrate among few companies, while inference becomes commoditized. New AI tools include Suno AI Personas for music creation, PromptQL for natural language querying over data, and Agent S for desktop task automation. Humor was shared about Python environment upgrades.
not much happened this weekend
claude-3.5-sonnet llama-3 llama-3-8b notebookllama min-omni-2 moondream openai anthropic hugging-face mistral-ai google-deepmind langchain deepmind microsoft pattern-recognition reinforcement-learning prompt-optimization text-to-speech model-optimization tensor-parallelism hyperparameters multimodal modal-alignment multimodal-fine-tuning ai-productivity privacy generative-ai rag retrieval-augmentation enterprise-text-to-sql amanda-askell philschmid stasbekman francois-fleuret mervenoyann reach_vb dzhng aravsrinivas sama lateinteraction andrew-y-ng bindureddy jerryjliu0
Moondream, a 1.6b vision language model, secured seed funding, highlighting a trend in moon-themed tiny models alongside Moonshine (27-61m ASR model). Claude 3.5 Sonnet was used for AI Twitter recaps. Discussions included pattern recognition vs. intelligence in LLMs, reinforcement learning for prompt optimization, and NotebookLlama, an open-source NotebookLM variant using LLaMA models for tasks like text-to-speech. Advances in model optimization with async-TP in PyTorch for tensor parallelism and hyperparameter tuning were noted. Mini-Omni 2 demonstrated multimodal capabilities across image, audio, and text for voice conversations with emphasis on modal alignment and multimodal fine-tuning. AI productivity tools like an AI email writer and LlamaCloud-based research assistants were introduced. Emphasis on practical skill development and privacy-conscious AI tool usage with Llama3-8B was highlighted. Generative AI tools such as #AIPythonforBeginners and GenAI Agents with LangGraph were shared. Business insights covered rapid execution in AI product development and emerging AI-related job roles. Challenges in enterprise-grade text-to-SQL and advanced retrieval methods were discussed with tutorials on RAG applications using LangChain and MongoDB.
not much happened today
llama-3.1-nemotron-70b golden-gate-claude embed-3 liquid-ai anthropic cohere openai meta-ai-fair nvidia perplexity-ai langchain kestra ostrisai llamaindex feature-steering social-bias multimodality model-optimization workflow-orchestration inference-speed event-driven-workflows knowledge-backed-agents economic-impact ai-national-security trust-dynamics sam-altman lmarena_ai aravsrinivas svpino richardmcngo ajeya_cotra tamaybes danhendrycks jerryjliu0
Liquid AI held a launch event introducing new foundation models. Anthropic shared follow-up research on social bias and feature steering with their "Golden Gate Claude" feature. Cohere released multimodal Embed 3 embeddings models following Aya Expanse. There was misinformation about GPT-5/Orion debunked by Sam Altman. Meta AI FAIR announced Open Materials 2024 with new models and datasets for inorganic materials discovery using the EquiformerV2 architecture. Anthropic AI demonstrated feature steering to balance social bias and model capabilities. NVIDIA's Llama-3.1-Nemotron-70B ranked highly on the Arena leaderboard with style control. Perplexity AI expanded to 100M weekly queries with new finance and reasoning modes. LangChain emphasized real application integration with interactive frame interpolation. Kestra highlighted scalable event-driven workflows with open-source YAML-based orchestration. OpenFLUX optimized inference speed by doubling it through guidance LoRA training. Discussions on AI safety included trust dynamics between humans and AI, economic impacts of AI automation, and the White House AI National Security memo addressing cyber and biological risks. LlamaIndex showcased knowledge-backed agents for enhanced AI applications.
not much happened today
claude-3.5-sonnet claude-3.5-haiku o1-preview mochi-1 stable-diffusion-3.5 embed-3 kerashub differential-transformer anthropic openai cohere microsoft computer-use coding-performance video-generation fine-tuning multimodality transformers attention-mechanisms model-optimization alexalbert fchollet rasbt
Anthropic released upgraded Claude 3.5 Sonnet and Claude 3.5 Haiku models featuring a new computer use capability that allows interaction with computer interfaces via screenshots and actions like mouse movement and typing. The Claude 3.5 Sonnet achieved state-of-the-art coding performance on SWE-bench Verified with a 49% score, surpassing OpenAI's o1-preview. Anthropic focuses on teaching general computer skills rather than task-specific tools, with expected rapid improvements. Other releases include Mochi 1, an open-source video generation model, Stable Diffusion 3.5 with Large and Medium variants, and Embed 3 by Cohere, a multimodal embedding model for text and image search. KerasHub was launched by François Chollet, unifying KerasNLP and KerasCV with 37 pretrained models. Microsoft introduced the Differential Transformer to reduce attention noise via differential attention maps, and research on transformer attention layers was shared by Rasbt.
DocETL: Agentic Query Rewriting and Evaluation for Complex Document Processing
bitnet-b1.58 llama-3.1-nemotron-70b-instruct gpt-4o claude-3.5-sonnet uc-berkeley deepmind openai microsoft nvidia archetype-ai boston-dynamics toyota-research google adobe openai mistral tesla meta-ai-fair model-optimization on-device-ai fine-tuning large-corpus-processing gpu-acceleration frameworks model-benchmarking rohanpaul_ai adcock_brett david-patterson
UC Berkeley's EPIC lab introduces innovative LLM data operators with projects like LOTUS and DocETL, focusing on effective programming and computation over large data corpora. This approach contrasts GPU-rich big labs like Deepmind and OpenAI with GPU-poor compound AI systems. Microsoft open-sourced BitNet b1.58, a 1-bit ternary parameter LLM enabling 4-20x faster training and on-device inference at human reading speeds. Nvidia released Llama-3.1-Nemotron-70B-Instruct, a fine-tuned open-source model outperforming GPT-4o and Claude-3.5-sonnet. These developments highlight advances in model-optimization, on-device-ai, and fine-tuning.
DeepSeek Janus and Meta SpiRit-LM: Decoupled Image and Expressive Voice Omnimodality
nemotron-70b claude claude-3.5-sonnet gpt-4o deepseek meta-ai-fair wandb nvidia anthropic hugging-face perplexity-ai multimodality image-generation speech-synthesis fine-tuning model-merging benchmarking open-source model-optimization reinforcement-learning bindureddy aravsrinivas danielhanchen clementdelangue cwolferesearch
DeepSeek Janus and Meta SpiRit-LM are two notable multimodality AI models recently released, showcasing advances in image generation and speech synthesis respectively. DeepSeek Janus separates vision encoders for image understanding and generation, achieving better results in both tasks. Meta's SpiRit-LM introduces an expressive speech and writing model generating pitch and style units, improving over standard TTS. Additionally, W&B Weave offers comprehensive LLM observability and multimodality fine-tuning tools. Industry updates include Nvidia's Nemotron 70b model underperforming, Meta open-sourcing Movie Gen Bench for media generation benchmarking, Perplexity launching internal search with multi-step reasoning, and Anthropic updating Claude apps. Open source progress includes Hugging Face's gradient accumulation fix in transformers and advocacy for open source AI to prevent Big Tech dominance. "Model merging for combining skills of multiple models" is also highlighted.
not much happened today
llama mistral openai decagon sierra togethercompute vertical-saas funding protein-structure-prediction lora self-supervised-learning model-optimization neural-architecture-search model-evaluation ethics transformers multi-agent-systems long-context mira-murati demis-hassabis clement-delangue john-o-whitaker yann-lecun francois-chollet ajeya-cotra rohan-paul adcock-brett
Vertical SaaS agents are gaining rapid consensus as the future of AI applications, highlighted by Decagon's $100m funding and Sierra's $4b round. OpenAI alumni are actively raising venture capital and forming new startups, intensifying competition in the AI market. Demis Hassabis celebrated the Nobel Prize recognition for AlphaFold2, a breakthrough in protein structure prediction. Advances in AI models include techniques like LoRA projectors and annealing on high-quality data, while discussions emphasize the need for high-bandwidth sensory inputs beyond language for common sense learning. New methods like LoLCATs aim to optimize transformer models such as Llama and Mistral for efficiency. Ethical concerns about AI agents performing harmful tasks remain under investigation. The AI community continues to explore model evaluation challenges and optimization frameworks like LPZero for neural architecture search.
Not much technical happened today
whisper-v3-turbo llama-3 llamaindex openai poolside liquidai perplexity-ai meta-ai-fair cohere fujitsu mixture-of-experts context-windows model-optimization fine-tuning quantization model-training alignment synthetic-data model-architecture agentic-ai nick-turley arav-srinivas francois-fleuret finbarr-timbers lewtun francois-chollet jerry-j-liu mmitchell-ai jxnlco
OpenAI announced raising $6.6B in new funding at a $157B valuation, with ChatGPT reaching 250M weekly active users. Poolside raised $500M to advance AGI development. LiquidAI introduced three new MoE models (1B, 3B, 40B) with a 32k context window and efficient token handling. OpenAI released Whisper V3 Turbo, an open-source multilingual model with significant speed improvements. Meta AI FAIR is hiring research interns focusing on LLM reasoning, alignment, synthetic data, and novel architectures. Cohere partnered with Fujitsu to launch Takane, a custom Japanese model. Technical discussions included challenges in LoRA fine-tuning, float8 quantization in Keras, and new tools like create-llama for agent templates. Industry commentary raised concerns about AI development priorities and highlighted freelancing opportunities in AI.
not much happened today
llama-3-2 llama-3 gemma-2 phi-3-5-mini claude-3-haiku gpt-4o-mini molmo gemini-1.5 gemini meta-ai-fair openai allenai google-deepmind multimodality model-optimization benchmarks ai-safety model-distillation pruning adapter-layers open-source-models performance context-windows mira-murati demis-hassabis ylecun sama
Meta AI released Llama 3.2 models including 1B, 3B text-only and 11B, 90B vision variants with 128K token context length and adapter layers for image-text integration. These models outperform competitors like Gemma 2 and Phi 3.5-mini, and are supported on major platforms including AWS, Azure, and Google Cloud. OpenAI CTO Mira Murati announced her departure. Allen AI released Molmo, an open-source multimodal model family outperforming proprietary systems. Google improved Gemini 1.5 with Flash and Pro models. Meta showcased Project Orion AR glasses and hinted at a Quest 3S priced at $300. Discussions covered new benchmarks for multimodal models, model optimization, and AI safety and alignment.
Llama 3.2: On-device 1B/3B, and Multimodal 11B/90B (with AI2 Molmo kicker)
llama-3-2 llama-3-1 claude-3-haiku gpt-4o-mini molmo-72b molmo-7b gemma-2 phi-3-5 llama-3-2-vision llama-3-2-3b llama-3-2-20b meta-ai-fair ai2 qualcomm mediatek arm ollama together-ai fireworks-ai weights-biases cohere weaviate multimodality vision context-windows quantization model-release tokenization model-performance model-optimization rag model-training instruction-following mira-murati daniel-han
Meta released Llama 3.2 with new multimodal versions including 3B and 20B vision adapters on a frozen Llama 3.1, showing competitive performance against Claude Haiku and GPT-4o-mini. AI2 launched multimodal Molmo 72B and 7B models outperforming Llama 3.2 in vision tasks. Meta also introduced new 128k-context 1B and 3B models competing with Gemma 2 and Phi 3.5, with collaborations hinted with Qualcomm, Mediatek, and Arm for on-device AI. The release includes a 9 trillion token count for Llama 1B and 3B. Partner launches include Ollama, Together AI offering free 11B model access, and Fireworks AI. Additionally, a new RAG++ course from Weights & Biases, Cohere, and Weaviate offers systematic evaluation and deployment guidance for retrieval-augmented generation systems based on extensive production experience.
a calm before the storm
o1 o1-mini qwen2.5 gpt-4 llama-2-70b llama-7b anthropic openai alibaba microsoft blackrock groq aramco disney eth-zurich pudu-robotics slack long-context kv-cache-quantization diffusion-models reinforcement-learning robotics ai-integration multilinguality model-benchmarking model-performance model-optimization adcock_brett philschmid rohanpaul_ai jvnixon kateclarktweets sama
Anthropic is raising funds at a valuation up to $40 billion ahead of anticipated major releases. OpenAI launched new reasoning models o1 and o1-mini, with increased rate limits and a multilingual MMLU benchmark. Alibaba released the open-source Qwen2.5 model supporting 29+ languages, showing competitive performance to gpt-4 at lower cost. Microsoft and Blackrock plan to invest $30 billion in AI data centers, with Groq partnering with Aramco to build the world's largest AI inference center. Robotics advances include Disney Research and ETH Zurich's diffusion-based motion generation for robots and Pudu Robotics' semi-humanoid robot. Slack and Microsoft introduced AI-powered agents integrated into their platforms. Research highlights include long-context scaling for llama-2-70b using Dual Chunk Attention and KV cache quantization enabling 1 million token context on llama-7b models.
Cerebras Inference: Faster, Better, AND Cheaper
llama-3.1-8b llama-3.1-70b gemini-1.5-flash gemini-1.5-pro cogvideox-5b mamba-2 rene-1.3b llama-3.1 gemini-1.5 claude groq cerebras cursor google-deepmind anthropic inference-speed wafer-scale-chips prompt-caching model-merging benchmarking open-source-models code-editing model-optimization jeremyphoward sam-altman nat-friedman daniel-gross swyx
Groq led early 2024 with superfast LLM inference speeds, achieving ~450 tokens/sec for Mixtral 8x7B and 240 tokens/sec for Llama 2 70B. Cursor introduced a specialized code edit model hitting 1000 tokens/sec. Now, Cerebras claims the fastest inference with their wafer-scale chips, running Llama3.1-8b at 1800 tokens/sec and Llama3.1-70B at 450 tokens/sec at full precision, with competitive pricing and a generous free tier. Google's Gemini 1.5 models showed significant benchmark improvements, especially Gemini-1.5-Flash and Gemini-1.5-Pro. New open-source models like CogVideoX-5B and Mamba-2 (Rene 1.3B) were released, optimized for consumer hardware. Anthropic's Claude now supports prompt caching, improving speed and cost efficiency. "Cerebras Inference runs Llama3.1 20x faster than GPU solutions at 1/5 the price."
Ideogram 2 + Berkeley Function Calling Leaderboard V2
llama-3-70b gpt-4 phi-3.5 functionary-llama-3-70b llama-3 ideogram midjourney berkeley openai hugging-face microsoft meta-ai-fair baseten kai claude functionary function-calling benchmarking image-generation model-optimization vision multimodality model-performance fine-tuning context-windows cybersecurity code-analysis ai-assisted-development
Ideogram returns with a new image generation model featuring color palette control, a fully controllable API, and an iOS app, reaching a milestone of 1 billion images created. Meanwhile, Midjourney released a Web UI but still lacks an API. In function calling, the Berkeley Function Calling Leaderboard (BFCL) updated to BFCL V2 • Live, adding 2251 live, user-contributed function documentation and queries to improve evaluation quality. GPT-4 leads the leaderboard, but the open-source Functionary Llama 3-70B finetune from Kai surpasses Claude. On AI model releases, Microsoft launched three Phi-3.5 models with impressive reasoning and context window capabilities, while Meta AI FAIR introduced UniBench, a unified benchmark suite for over 50 vision-language model tasks. Baseten improved Llama 3 inference speed by up to 122% using Medusa. A new cybersecurity benchmark, Cyberbench, featuring 40 CTF tasks, was released. Additionally, Codegen was introduced as a tool for programmatic codebase analysis and AI-assisted development. "Multiple functions > parallel functions" was highlighted as a key insight in function calling.
The DSPy Roadmap
dspy litel-lm gemini chatgpt-4o grok-2 hermes-3 databricks mit google openai x-ai nous-research astribot apple sakana-ai model-optimization fine-tuning optimizers interactive-optimization robotics autonomous-systems voice image-generation open-source-models scientific-research streaming caching omar-khattab giffmana
Omar Khattab announced joining Databricks before his MIT professorship and outlined the roadmap for DSPy 2.5 and 3.0+, focusing on improving core components like LMs, signatures, optimizers, and assertions with features such as adopting LiteLLM to reduce code and enhance caching and streaming. The roadmap also includes developing more accurate, cost-effective optimizers, building tutorials, and enabling interactive optimization tracking. On AI Twitter, Google launched Gemini Live, a mobile conversational AI with voice and 10 voices, alongside Pixel Buds Pro 2 with a custom Tensor A1 chip. OpenAI updated ChatGPT-4o, reclaiming the top spot on LMSYS Arena. xAI released Grok-2 in beta, achieving SOTA in image generation with FLUX 1. Nous Research released open-source Hermes 3 models in 8B, 70B, and 405B sizes, with the 405B model achieving SOTA. Robotics updates include Astribot's humanoid robot and Apple's tabletop robot with Siri voice commands. Sakana AI introduced "The AI Scientist," an autonomous AI research system.
DataComp-LM: the best open-data 7B model/benchmark/dataset
mistral-nemo-12b gpt-4o-mini deepseek-v2-0628 mistral-7b llama-3 gemma-2 qwen-2 datacomp hugging-face openai nvidia mistral-ai deepseek dataset-design scaling-laws model-benchmarking model-performance fine-tuning multilinguality function-calling context-windows open-source-models model-optimization cost-efficiency benchmarking sam-altman guillaume-lample philschmid miramurati
DataComp team released a competitive 7B open data language model trained on only 2.5T tokens from the massive DCLM-POOL dataset of 240 trillion tokens, showing superior scaling trends compared to FineWeb. OpenAI launched GPT-4o mini, a cost-effective model with 82% MMLU and performance near GPT-4-Turbo, aimed at developers for broad applications. NVIDIA and Mistral jointly released the Mistral NeMo 12B model featuring a 128k token context window, FP8 checkpoint, multilingual support, and Apache 2.0 licensing. DeepSeek announced DeepSeek-V2-0628 as the top open-source model on the LMSYS Chatbot Arena leaderboard with strong rankings in coding, math, and hard prompts. This news highlights advances in dataset design, model efficiency, and open-source contributions in the AI community.
Mini, Nemo, Turbo, Lite - Smol models go brrr (GPT4o-mini version)
gpt-4o-mini deepseek-v2-0628 mistral-nemo llama-8b openai deepseek-ai mistral-ai nvidia meta-ai-fair hugging-face langchain keras cost-efficiency context-windows open-source benchmarking neural-networks model-optimization text-generation fine-tuning developer-tools gpu-support parallelization cuda-integration multilinguality long-context article-generation liang-wenfeng
OpenAI launched the GPT-4o Mini, a cost-efficient small model priced at $0.15 per million input tokens and $0.60 per million output tokens, aiming to replace GPT-3.5 Turbo with enhanced intelligence but some performance limitations. DeepSeek open-sourced DeepSeek-V2-0628, topping the LMSYS Chatbot Arena Leaderboard and emphasizing their commitment to contributing to the AI ecosystem. Mistral AI and NVIDIA released the Mistral NeMo, a 12B parameter multilingual model with a record 128k token context window under an Apache 2.0 license, sparking debates on benchmarking accuracy against models like Meta Llama 8B. Research breakthroughs include the TextGrad framework for optimizing compound AI systems via textual feedback differentiation and the STORM system improving article writing by 25% through simulating diverse perspectives and addressing source bias. Developer tooling trends highlight LangChain's evolving context-aware reasoning applications and the Modular ecosystem's new official GPU support, including discussions on Mojo and Keras 3.0 integration.
Qdrant's BM42: "Please don't trust us"
claude-3.5-sonnet gemma-2 nano-llava-1.5 qdrant cohere stripe anthropic hugging-face stablequan_ai semantic-search benchmarking dataset-quality model-evaluation model-optimization vision fine-tuning context-windows nils-reimers jeremyphoward hamelhusain rohanpaul_ai
Qdrant attempted to replace BM25 and SPLADE with a new method called "BM42" combining transformer attention and collection-wide statistics for semantic and keyword search, but their evaluation using the Quora dataset was flawed. Nils Reimers from Cohere reran BM42 on better datasets and found it underperformed. Qdrant acknowledged the errors but still ran a suboptimal BM25 implementation. This highlights the importance of dataset choice and evaluation sanity checks in search model claims. Additionally, Stripe faced criticism for AI/ML model failures causing account and payment issues, prompting calls for alternatives. Anthropic revealed that Claude 3.5 Sonnet suppresses some answer parts with backend tags, sparking debate. Gemma 2 model optimizations allow 2x faster fine-tuning with 63% less memory and longer context windows, running up to 34B parameters on consumer GPUs. nanoLLaVA-1.5 was announced as a compact 1B parameter vision model with significant improvements.
RouteLLM: RIP Martian? (Plus: AINews Structured Summaries update)
gpt-4 gemma-2-27b gemma-2-9b lmsys openai llm-routing cost-efficiency model-performance model-optimization data-augmentation syntax-based-routing mixture-of-experts inference-throughput software-2.0 computer-vision karpathy bindureddy armand-joulin
LMSys introduces RouteLLM, an open-source router framework trained on preference data from Chatbot Arena, achieving cost reductions over 85% on MT Bench, 45% on MMLU, and 35% on GSM8K while maintaining 95% of GPT-4's performance. This approach surpasses previous task-specific routing by using syntax-based Mixture of Experts (MoE) routing and data augmentation, beating commercial solutions by 40%. The update highlights advances in LLM routing, cost-efficiency, and model performance optimization across multiple models rather than single-model or MoE-level improvements. Additionally, the AI Twitter recap notes the Gemma 2 model family as a top open model, the Block Transformer architecture for improved inference throughput, and a proposal for a fully Software 2.0 computer vision system by karpathy.
That GPT-4o Demo
gpt-4o gemma-2 meta-code-llama openai google-deepmind meta-ai-fair voice-generation ocr screen-sharing vision code-understanding model-customization efficiency textual-intelligence multimodal-agents sft distillation rlhf model-merging model-optimization safety romain-huet fchollet
Romain Huet demonstrated an unreleased version of GPT-4o on ChatGPT Desktop showcasing capabilities like low latency voice generation, whisper tone moderation, camera mode streaming video to GPT-4o, rapid OCR, screen sharing with ChatGPT for programming help, clipboard reading, and vision-based code conversation. OpenAI's four investment areas highlighted include textual intelligence, efficiency/cost, model customization, and multimodal agents. Google DeepMind released Gemma 2 models in 9B and 27B sizes trained on 8T and 13T tokens respectively, using SFT, distillation, RLHF, and model merging, optimized for TPUv5e with strong performance and safety measures. Meta AI announced the Meta LLM Compiler built on Meta Code Llama with enhanced code optimization and compiler features.
Gemma 2: The Open Model for Everyone
gemma-2 qwen-72b mixtral-8x22b-instruct claude-3.5-sonnet google-deepmind alibaba mistral-ai anthropic knowledge-distillation attention-mechanisms multilingual-models multimodality model-training model-optimization memory-optimization fine-tuning kathleen-kenealy daniel-han
Gemma 2, a 27B parameter model from google-deepmind, was released with innovations like 1:1 local-global attention alternation and logit soft-capping, leveraging knowledge distillation to train smaller models on over 50× the compute-optimal token quantity. The model supports multilingual and multimodal capabilities, with fine-tuning success on over 200 Indic language variants. The Open LLM Leaderboard highlights alibaba's Qwen 72B as the top model, with mistral-ai's Mixtral-8x22B-Instruct also ranking highly. Anthropic launched Claude 3.5 Sonnet, improving intelligence at mid-tier cost and speed. Research on eliminating matrix multiplication in LLMs promises significant memory savings without performance loss. Kathleen Kenealy and Daniel Han provided insights on Gemma 2's tokenizer and attention scaling respectively.
Claude Crushes Code - 92% HumanEval and Claude.ai Artifacts
claude-3.5-sonnet claude-3-opus gpt-4o anthropic openai cognition benchmarking model-performance coding model-optimization fine-tuning instruction-following model-efficiency model-release api performance-optimization alex-albert
Claude 3.5 Sonnet, released by Anthropic, is positioned as a Pareto improvement over Claude 3 Opus, operating at twice the speed and costing one-fifth as much. It achieves state-of-the-art results on benchmarks like GPQA, MMLU, and HumanEval, surpassing even GPT-4o and Claude 3 Opus on vision tasks. The model demonstrates significant advances in coding capabilities, passing 64% of test cases compared to 38% for Claude 3 Opus, and is capable of autonomously fixing pull requests. Anthropic also introduced the Artifacts feature, enabling users to interact with AI-generated content such as code snippets and documents in a dynamic workspace, similar to OpenAI's Code Interpreter. This release highlights improvements in performance, cost-efficiency, and coding proficiency, signaling a growing role for LLMs in software development.
Talaria: Apple's new MLOps Superweapon
gemma mixtral phi dbrx apple google mistral-ai microsoft mosaic quantization on-device-ai adapter-models model-optimization model-latency lossless-quantization low-bit-palletization token-generation model-benchmarking human-evaluation craig-federighi andrej-karpathy
Apple Intelligence introduces a small (~3B parameters) on-device model and a larger server model running on Apple Silicon with Private Cloud Compute, aiming to surpass Google Gemma, Mistral Mixtral, Microsoft Phi, and Mosaic DBRX. The on-device model features a novel lossless quantization strategy using mixed 2-bit and 4-bit LoRA adapters averaging 3.5 bits-per-weight, enabling dynamic adapter hot-swapping and efficient memory management. Apple credits the Talaria tool for optimizing quantization and model latency, achieving about 0.6 ms time-to-first-token latency and 30 tokens per second generation rate on iPhone 15 Pro. Apple focuses on an "adapter for everything" strategy with initial deployment on SiriKit and App Intents. Performance benchmarks rely on human graders, emphasizing consumer-level adequacy over academic dominance. The Apple ML blog also mentions an Xcode code-focused model and a diffusion model for Genmoji.
Not much happened today
gemini-1.5-flashmodel gemini-pro mixtral mamba-2 phi-3-medium phi-3-small gpt-3.5-turbo-0613 llama-3-8b llama-2-70b mistral-finetune twelve-labs livekit groq openai nea nvidia lmsys mistral-ai model-performance prompt-engineering data-curation ai-safety model-benchmarking model-optimization training sequence-models state-space-models daniel-kokotajlo rohanpaul_ai _arohan_ tri_dao _albertgu _philschmid sarahcat21 hamelhusain jachiam0 willdepue teknium1
Twelve Labs raised $50m in Series A funding co-led by NEA and NVIDIA's NVentures to advance multimodal AI. Livekit secured $22m in funding. Groq announced running at 800k tokens/second. OpenAI saw a resignation from Daniel Kokotajlo. Twitter users highlighted Gemini 1.5 FlashModel for high performance at low cost and Gemini Pro ranking #2 in Japanese language tasks. Mixtral models can run up to 8x faster on NVIDIA RTX GPUs using TensorRT-LLM. Mamba-2 model architecture introduces state space duality for larger states and faster training, outperforming previous models. Phi-3 Medium (14B) and Small (7B) models benchmark near GPT-3.5-Turbo-0613 and Llama 3 8B. Prompt engineering is emphasized for unlocking LLM capabilities. Data quality is critical for model performance, with upcoming masterclasses on data curation. Discussions on AI safety include a Frontier AI lab employee letter advocating whistleblower protections and debates on aligning AI to user intent versus broader humanity interests.
Skyfall
gemini-1.5-pro gemini-1.5-flash yi-1.5 kosmos-2.5 paligemma falcon-2 deepseek-v2 hunyuan-dit gemini-1.5 gemini-1.5-flash yi-1.5 google-deepmind yi-ai microsoft hugging-face langchain maven multimodality mixture-of-experts transformer model-optimization long-context model-performance model-inference fine-tuning local-ai scaling-laws causal-models hallucination-detection model-distillation model-efficiency hamel-husain dan-becker clement-delangue philschmid osanseviero arankomatsuzaki jason-wei rohanpaul_ai
Between 5/17 and 5/20/2024, key AI updates include Google DeepMind's Gemini 1.5 Pro and Flash models, featuring sparse multimodal MoE architecture with up to 10M context and a dense Transformer decoder that is 3x faster and 10x cheaper. Yi AI released Yi-1.5 models with extended context windows of 32K and 16K tokens. Other notable releases include Kosmos 2.5 (Microsoft), PaliGemma (Google), Falcon 2, DeepSeek v2 lite, and HunyuanDiT diffusion model. Research highlights feature an Observational Scaling Laws paper predicting model performance across families, a Layer-Condensed KV Cache technique boosting inference throughput by up to 26×, and the SUPRA method converting LLMs into RNNs for reduced compute costs. Hugging Face expanded local AI capabilities enabling on-device AI without cloud dependency. LangChain updated its v0.2 release with improved documentation. The community also welcomed a new LLM Finetuning Discord by Hamel Husain and Dan Becker for Maven course users. "Hugging Face is profitable, or close to profitable," enabling $10 million in free shared GPUs for developers.
Google I/O in 60 seconds
gemini-1.5-pro gemini-flash gemini-ultra gemini-pro gemini-nano gemma-2 llama-3-70b paligemma imagen-3 veo google google-deepmind youtube tokenization model-performance fine-tuning vision multimodality model-release model-training model-optimization ai-integration image-generation watermarking hardware-optimization voice video-understanding
Google announced updates to the Gemini model family, including Gemini 1.5 Pro with 2 million token support, and the new Gemini Flash model optimized for speed with 1 million token capacity. The Gemini suite now includes Ultra, Pro, Flash, and Nano models, with Gemini Nano integrated into Chrome 126. Additional Gemini features include Gemini Gems (custom GPTs), Gemini Live for voice conversations, and Project Astra, a live video understanding assistant. The Gemma model family was updated with Gemma 2 at 27B parameters, offering near-llama-3-70b performance at half the size, plus PaliGemma, a vision-language open model inspired by PaLI-3. Other launches include DeepMind's Veo, Imagen 3 for photorealistic image generation, and a Music AI Sandbox collaboration with YouTube. SynthID watermarking now extends to text, images, audio, and video. The Trillium TPUv6 codename was revealed. Google also integrated AI across its product suite including Workspace, Email, Docs, Sheets, Photos, Search, and Lens. "The world awaits Apple's answer."
GPT-4o: the new SOTA-EVERYTHING Frontier model (GPT4O version)
gpt-4o gpt-4-turbo openai lmsys multion adept multimodality vision speech-recognition tokenization real-time-processing coding model-performance model-optimization desktop-agents sama gdb
OpenAI has released GPT-4o, a new multimodal model capable of reasoning across text, audio, and video in real time with low latency (~300ms). It features voice and vision capabilities, improved non-English language performance with an expanded 200k vocabulary tokenizer, and is available to all ChatGPT users including free plans. GPT-4o is half the price and twice as fast as GPT-4-turbo with 5x rate limits. The model supports real-time voice and video input/output and shows strong coding capabilities. The release includes a new desktop app that can read screen and clipboard history, challenging existing desktop agent startups. The announcement was accompanied by demos including image generation and 3D object handling, with OpenAI achieving state-of-the-art performance in ASR and vision tasks. The update was widely discussed on social media, with comparisons to GPT-4T highlighting GPT-4o's speed and versatility. "GPT-4o is smart, fast, natively multimodal, and a step towards more natural human-computer interaction" and "extremely versatile and fun to play with".
A quiet weekend
llama-3 dolphin-2.9 pixart-sigma llama-3-70b microsoft coca-cola uber lmsys nous-research mistral-ai ar-interfaces transformers algorithmic-tasks turing-test graph-algorithms embeddings generative-ai model-optimization llm-inference quantization model-deployment yann-lecun
Yann LeCun predicts a shift to AR interfaces with AI assistants in 10-15 years, moving away from smartphones. The Dolphin-2.9 model based on Llama-3 was released, improving quality issues. PixArt Sigma, a 0.6B parameter model, achieves Stable Diffusion 3.0 level performance with complete prompt adherence and local usability. Research shows transformers can use meaningless filler tokens for algorithmic tasks with dense supervision. AI-generated restaurant reviews can pass the Turing test, fooling humans and AI detectors. Uber uses graph algorithms and learned embeddings for ETA prediction. Coca-Cola and Microsoft announced a 5-year AI partnership to accelerate cloud and generative AI initiatives. The Llama-3 70B model can run on a single 4GB GPU using AirLLM optimization without quantization but is slow. Mistral.rs is introduced as a fast LLM inference platform with quantization and OpenAI API compatibility. Only 5% of LLMs make it from prototype to production due to challenges, especially in enterprise. EXL2 and GGUF quantization methods for Llama models show similar perplexity vs model size, with Llama-3 and Llama-2 degrading more under quantization compared to full precision.
Anime pfp anon eclipses $10k A::B prompting challenge
command-r-plus-104b stable-diffusion-1.5 openai ollama huggingface quantization model-optimization streaming prompt-engineering self-prompting image-composition character-lora-training model-size open-source-licenses memes humor victor-taelin futuristfrog
Victor Taelin issued a $10k challenge to GPT models, initially achieving only 10% success with state-of-the-art models, but community efforts surpassed 90% success within 48 hours, highlighting GPT capabilities and common skill gaps. In Reddit AI communities, Command R Plus (104B) is running quantized on M2 Max hardware via Ollama and llama.cpp forks, with GGUF quantizations released on Huggingface. Streaming text-to-video generation is now available through the st2v GitHub repo. WD Tagger v3 was released for mass auto-captioning datasets with a WebUI. Lesser-known prompting techniques like self-tagging and generational frameworks produced thought-provoking outputs in OpenAI discussions, including experiments with self-evolving system prompts. Stable Diffusion users discussed image composition importance for training character LoRAs and best checkpoints for video game character generation. Discussions also covered scarcity of 5B parameter models and open(ish) licenses for open source AI. Memes included jokes about ChatGPT and Gemini training data differences.
Evals-based AI Engineering
jamba bamboo qwen-1.5-moe grok-1.5 llama2-7b openai mistral-ai x-ai llamaindex evaluation fine-tuning prompt-engineering voice-cloning quantization model-optimization code-generation context-windows hamel-husain alec-radford
Hamel Husain emphasizes the importance of comprehensive evals in AI product development, highlighting evaluation, debugging, and behavior change as key iterative steps. OpenAI released a voice engine demo showcasing advanced voice cloning from small samples, raising safety concerns. Reddit discussions introduced new models like Jamba (hybrid Transformer-SSM with MoE), Bamboo (7B LLM with high sparsity based on Mistral), Qwen1.5-MoE (efficient parameter activation), and Grok 1.5 (128k context length, surpassing GPT-4 in code generation). Advances in quantization include 1-bit Llama2-7B models outperforming full precision and the QLLM quantization toolbox supporting GPTQ/AWQ/HQQ methods.
Jamba: Mixture of Architectures dethrones Mixtral
jamba dbrx mixtral animatediff fastsd sdxs512-0.9 b-lora supir ai21-labs databricks together-ai hugging-face midjourney mixture-of-experts model-architecture context-windows model-optimization fine-tuning image-generation video-generation cpu-optimization style-content-separation high-resolution-upscaling
AI21 labs released Jamba, a 52B parameter MoE model with 256K context length and open weights under Apache 2.0 license, optimized for single A100 GPU performance. It features a unique blocks-and-layers architecture combining transformer and MoE layers, competing with models like Mixtral. Meanwhile, Databricks introduced DBRX, a 36B active parameter MoE model trained on 12T tokens, noted as a new standard for open LLMs. In image generation, advancements include Animatediff for video-quality image generation and FastSD CPU v1.0.0 beta 28 enabling ultra-fast image generation on CPUs. Other innovations involve style-content separation using B-LoRA and improvements in high-resolution image upscaling with SUPIR.
The Era of 1-bit LLMs
bitnet-b1.58 hugging-face quantization model-optimization energy-efficiency fine-tuning robotics multimodality ai-security ethics humor swyx levelsio gdb npew _akhaliq osanseviero mmitchell_ai deliprao nearcyan clementdelangue
The Era of 1-bit LLMs research, including the BitNet b1.58 model, introduces a ternary parameter approach that matches full-precision Transformer LLMs in performance while drastically reducing energy costs by 38x. This innovation promises new scaling laws and hardware designs optimized for 1-bit LLMs. Discussions on AI Twitter highlight advances in AGI societal impact, robotics with multimodal models, fine-tuning techniques like ResLoRA, and AI security efforts at Hugging Face. Ethical considerations in generative AI and humor within the AI community are also prominent topics.
Welcome Interconnects and OpenRouter
mistral-large miqu mixtral gpt-4 mistral-7b mistral-ai openai perplexity-ai llamaindex qwen langchain model-comparison model-optimization quantization role-playing story-writing code-clarity ai-assisted-decompilation asynchronous-processing quantum-computing encoder-based-diffusion open-source hardware-experimentation rag-systems nathan-lambert alex-atallah
Discord communities analyzed 22 guilds, 349 channels, and 12885 messages revealing active discussions on model comparisons and optimizations involving Mistral AI, Miqu, and GGUF quantized models. Highlights include comparing Mistral Large with GPT-4, focusing on cost-effectiveness and performance, and exploring quantization techniques like GPTQ and QLORA to reduce VRAM usage. Advanced applications such as role-playing, story-writing, code clarity, and AI-assisted decompilation were emphasized, alongside development of tools like an asynchronous summarization script for Mistral 7b. The intersection of quantum computing and AI was discussed, including DARPA-funded projects and encoder-based diffusion techniques for image processing. Community efforts featured new Spanish LLM announcements, hardware experimentation, and open-source initiatives, with platforms like Perplexity AI and LlamaIndex noted for innovation and integration. Speculation about Mistral AI's open-source commitment and tools like R2R for rapid RAG deployment highlighted collaborative spirit.
Mistral Large disappoints
mistral-large mistral-small mixtral-8x7b gpt-4-turbo dreamgen-opus-v1 mistral-ai openai hugging-face benchmarking model-merging fine-tuning reinforcement-learning model-training tokenization model-optimization ai-assisted-decompilation performance cost-efficiency deception roleplay deep-speed dpo timotheeee1 cogbuji plasmator jsarnecki maldevide spottyluck mrjackspade
Mistral announced Mistral Large, a new language model achieving 81.2% accuracy on MMLU, trailing GPT-4 Turbo by about 5 percentage points on benchmarks. The community reception has been mixed, with skepticism about open sourcing and claims that Mistral Small outperforms the open Mixtral 8x7B. Discussions in the TheBloke Discord highlighted performance and cost-efficiency comparisons between Mistral Large and GPT-4 Turbo, technical challenges with DeepSpeed and DPOTrainer for training, advances in AI deception for roleplay characters using DreamGen Opus V1, and complexities in model merging using linear interpolation and PEFT methods. Enthusiasm for AI-assisted decompilation was also expressed, emphasizing the use of open-source projects for training data.
Karpathy emerges from stealth?
mistral-7b mixtral-8x7b zephyr-7b gpt-4 llama-2 intel mistral-ai audiogen thebloke tokenization quantization model-optimization fine-tuning model-merging computational-efficiency memory-optimization retrieval-augmented-generation multi-model-learning meta-reasoning dataset-sharing open-source ethical-ai community-collaboration andrej-karpathy
Andrej Karpathy released a comprehensive 2-hour tutorial on tokenization, detailing techniques up to GPT-4's tokenizer and noting the complexity of Llama 2 tokenization with SentencePiece. Discussions in AI Discord communities covered model optimization and efficiency, focusing on quantization of models like Mistral 7B and Zephyr-7B to reduce memory usage for consumer GPUs, including Intel's new weight-only quantization algorithm. Efforts to improve computational efficiency included selective augmentation reducing costs by 57.76% and memory token usage versus kNN for Transformers. Challenges in hardware compatibility and software issues were shared, alongside fine-tuning techniques such as LoRA and model merging. Innovative applications of LLMs in retrieval-augmented generation (RAG), multi-model learning, and meta-reasoning were explored. The community emphasized dataset sharing, open-source releases like SDXL VAE encoded datasets and Audiogen AI codecs, and ethical AI use with censorship and guardrails. Collaboration and resource sharing remain strong in these AI communities.
AI2 releases OLMo - the 4th open-everything LLM
olmo-1b olmo-7b olmo-65b miqu-70b mistral-medium distilbert-base-uncased ai2 allenai mistral-ai tsmc asml zeiss fine-tuning gpu-shortage embedding-chunking json-generation model-optimization reproducible-research self-correction vram-constraints programming-languages nathan-lambert lhc1921 mrdragonfox yashkhare_ gbourdin
AI2 is gaining attention in 2024 with its new OLMo models, including 1B and 7B sizes and a 65B model forthcoming, emphasizing open and reproducible research akin to Pythia. The Miqu-70B model, especially the Mistral Medium variant, is praised for self-correction and speed optimizations. Discussions in TheBloke Discord covered programming language preferences, VRAM constraints for large models, and fine-tuning experiments with Distilbert-base-uncased. The Mistral Discord highlighted challenges in the GPU shortage affecting semiconductor production involving TSMC, ASML, and Zeiss, debates on open-source versus proprietary models, and fine-tuning techniques including LoRA for low-resource languages. Community insights also touched on embedding chunking strategies and JSON output improvements.
Miqu confirmed to be an early Mistral-medium checkpoint
miqu-1-70b mistral-medium llama-2-70b-chat mixtral sqlcoder-70b codellama-70b bagelmistery-tour-v2 psyfighter-v2 mistral-ai hugging-face nous-research aiatmeta instruction-following sampling-methods fp16-quantization fine-tuning model-training context-length text-to-sql model-performance model-optimization intrstllrninja
Miqu, an open access model, scores 74 on MMLU and 84.5 on EQ-Bench, sparking debates about its performance compared to Mistral Medium. The CEO of Mistral confirmed these results. Discussions in the TheBloke Discord highlight Miqu's superiority in instruction-following and sampling methods like dynatemp and min-p. Developers also explore browser preferences and Discord UI themes. Role-playing with models like BagelMistery Tour v2 and Psyfighter v2 is popular, alongside technical talks on fp16 quantization of Miqu-1-70b. Training and fine-tuning tips for models like Unsloth and Mistral 7B are shared. In the Nous Research AI Discord, the Activation Beacon method is discussed for extending LLM context length from 4K to 400K tokens. SQLCoder-70B, fine-tuned on CodeLlama-70B, leads in text-to-SQL generation and is available on Hugging Face. The Miqu model also impresses with an 83.5 EQ-Bench score, fueling speculation about its capabilities.
RWKV "Eagle" v5: Your move, Mamba
rwkv-v5 mistral-7b miqu-1-70b mistral-medium llama-2 mistral-instruct-v0.2 mistral-tuna llama-2-13b kunoichi-dpo-v2-7b gpt-4 eleutherai mistral-ai hugging-face llamaindex nous-research rwkv lmsys fine-tuning multilinguality rotary-position-embedding model-optimization model-performance quantization speed-optimization prompt-engineering model-benchmarking reinforcement-learning andrej-karpathy
RWKV v5 Eagle was released with better-than-mistral-7b evaluation results, trading some English performance for multilingual capabilities. The mysterious miqu-1-70b model sparked debate about its origins, possibly a leak or distillation of Mistral Medium or a fine-tuned Llama 2. Discussions highlighted fine-tuning techniques, including the effectiveness of 1,000 high-quality prompts over larger mixed-quality datasets, and tools like Deepspeed, Axolotl, and QLoRA. The Nous Research AI community emphasized the impact of Rotary Position Embedding (RoPE) theta settings on LLM extrapolation, improving models like Mistral Instruct v0.2. Speed improvements in Mistral Tuna kernels reduced token processing costs, enhancing efficiency. The launch of Eagle 7B with 7.52B parameters showcased strong multilingual performance, surpassing other 7B class models.
GPT4Turbo A/B Test: gpt-4-0125-preview
gpt-4-turbo gpt-4-1106-preview gpt-3.5 llama-2-7b-chat tiny-llama mistral openai thebloke nous-research hugging-face multi-gpu-support model-optimization model-merging fine-tuning context-windows chatbot-personas api-performance text-transcription cost-considerations model-troubleshooting
OpenAI released a new GPT-4 Turbo version in January 2024, prompting natural experiments in summarization and discussions on API performance and cost trade-offs. The TheBloke Discord highlighted UnSloth's upcoming limited multi-GPU support for Google Colab beginners, AI models like Tiny Llama and Mistral running on Nintendo Switch, and advanced model merging techniques such as DARE and SLERP. The OpenAI Discord noted issues with GPT-4-1106-preview processing delays, troubleshooting GPT model errors, and transcription challenges with GPT-3.5 and GPT-4 Turbo. Nous Research AI focused on extending context windows, notably LLaMA-2-7B-Chat reaching 16,384 tokens, and fine-tuning alternatives like SelfExtend. Discussions also touched on chatbot persona creation, model configuration optimizations, and societal impacts of AI technology.
12/25/2023: Nous Hermes 2 Yi 34B for Christmas
nous-hermes-2 yi-34b nucleusx yayi-2 ferret teknim nous-research apple mixtral deepseek qwen huggingface wenge-technology quantization model-optimization throughput-metrics batch-processing parallel-decoding tensor-parallelization multimodality language-model-pretraining model-benchmarking teknium carsonpoole casper_ai pradeep1148 osanseviero metaldragon01
Teknium released Nous Hermes 2 on Yi 34B, positioning it as a top open model compared to Mixtral, DeepSeek, and Qwen. Apple introduced Ferret, a new open-source multimodal LLM. Discussions in the Nous Research AI Discord focused on AI model optimization and quantization techniques like AWQ, GPTQ, and AutoAWQ, with insights on proprietary optimization and throughput metrics. Additional highlights include the addition of NucleusX Model to transformers, a 30B model with 80 MMLU, and the YAYI 2 language model by Wenge Technology trained on 2.65 trillion tokens. "AutoAWQ outperforms vLLM up to batch size 8" was noted, and proprietary parallel decoding and tensor parallelization across GPUs were discussed for speed improvements.