All tags
Person: "yuchenj_uw"
not much happened today
gpt-5.6 claude-fable-5 openai model-stratification agentic-coding presentation benchmarking orchestration computer-use gui-automation reward-hacking instruction-following usage-limits model-costs reach_vb rasbt yuchenj_uw scaling01 simonw kimmonismus thsottiaux htihle teortaxestex mononofu omarsar0 hangsiin gdb mckbrando evi77ain
OpenAI rolled out GPT-5.6 featuring a new model stratification with tiers Luna / Terra / Sol and effort levels including Max and Ultra, introducing complex configuration options. The launch faced UX challenges with the ChatGPT Work / Codex split, prompting rapid corrective actions including usage-limit resets and UI improvements. Early benchmarks show GPT-5.6 excels in agentic coding, presentation, and science tasks, tying with Claude Fable 5 in Code Arena Frontend at about half the cost, and achieving a significant 500-point Elo gain in presentations. However, users noted instruction-following issues and concerns about jailbreakability. The major advancement is in orchestration and computer use, with Sol Ultra demonstrating strong planner and verifier capabilities, enabling high-throughput automation workflows. A notable operational challenge is the hidden cost explosion from spawned subagents inheriting premium settings, causing faster quota depletion.
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
claude-opus-4.7 gemini-3.1-pro gpt-5.4 claude-code codex anthropic openai agentic-ai model-benchmarking adaptive-reasoning cost-efficiency computer-use prototyping-tools code-generation model-performance software-integration claudeai yuchenj_uw kimmonismus skirano therundownai arena artificialanlys victortaelin emollick alexalbert__ theo scaling01 reach_vb kr0der hamelhusain mattrickard matvelloso gdb
Anthropic launched Claude Design, a prototyping tool powered by Claude Opus 4.7, targeting design workflows and competing with Figma and others. Benchmarks show Opus 4.7 leading in coding and text tasks, with improved efficiency and adaptive reasoning, though early user feedback noted some regressions and stability issues. Discussions highlighted its cost-efficiency and agentic capabilities compared to Gemini 3.1 Pro and GPT-5.4. Meanwhile, OpenAI's Codex updates introduced advanced computer-use features enabling fast, agentic control of desktop apps and enterprise software, signaling progress toward practical AGI-like agents.
not much happened today
Hermes Agent is gaining attention as a leading open agent stack with features like self-improving skills, persistent memory, and a self-improvement loop. Its new Manim skill enables generation of math/technical animations, expanding agent capabilities. The Hermes ecosystem is rapidly growing with GUI tools, WebUI, HUD updates, OAuth support, and integrations. An open training-data movement for agents is emerging, focusing on sharing reusable behavioral data and harness traces. Meanwhile, Anthropic's Claude Code faces distribution and policy challenges, with reports of restrictions and unreliability impacting third-party coding agents, highlighting issues with subscription economics for always-on agents. "Claude Code now errors if used to analyze Claude Code source" and "basically unusable" are key community sentiments.
not much happened today
claude-opus-4.6 capybara glm-5.1 qwen-3.5-14b qwen-27b qwen3.5-35b anthropic google zhipu model-scaling coding academic-reasoning cybersecurity quantization local-inference model-benchmarking inference-optimization model-performance agent-products scaling01 yuchenj_uw kimmonismus m1astra dejavucoder iscienceluvr gaoj0017
Anthropic is reportedly introducing a new AI model tier called Capybara, which is larger and more intelligent than Claude Opus 4.6, showing improved performance in coding, academic reasoning, and cybersecurity. The model is speculated to be around 10 trillion parameters, with Google potentially funding Anthropic's data center expansion. Meanwhile, Zhipu released GLM-5.1, advancing open coding models and narrowing the gap with closed models. Local inference economics are improving, highlighted by efficient deployments of Qwen 3.5 14B, Qwen 27B, and Qwen3.5-35B models with quantization techniques like TurboQuant vLLM. However, TurboQuant's benchmarking claims face criticism from researchers. Overall, the AI landscape shows aggressive scaling, local model deployment, and agent products gaining traction.
not much happened today
kimi-k2.5 claude-code cursor kimi fireworks anthropic langchain model-attribution fine-tuning reinforcement-learning open-source agent-products model-licensing software-integration product-differentiation clementdelangue leerob amanrsanger yuchenj_uw kimmonismus
Cursor's Composer 2, built on Kimi K2.5, sparked discussion over model attribution and licensing, highlighting a shift toward post-trained derivatives of open-source models with domain-specific fine-tuning and reinforcement learning. Claude Code is expanding into third-party tools like T3 Code and communication channels such as Telegram and Discord, while LangChain is evolving from orchestration to multi-agent products with offerings like Deep Agents/Open SWE and LangSmith Fleet. The discourse emphasizes the importance of clear base-model attribution, licensing compliance, and product differentiation through fine-tuning and user experience.
not much happened today
claude-code composer-2 cursor openai anthropic langchain cognition reinforcement-learning developer-tooling agent-systems agent-runtimes security credential-management multi-agent-systems model-training benchmarking software-engineering enterprise-ai kimmonismus mntruell theo ellev3n11 amanrsanger charliermarsh gdb yuchenj_uw neilhtennek simonw yuvalinthedeep lvwerra hrishioa
Cursor launched Composer 2, a frontier-class coding model with major cost reductions and strong benchmark scores like 61.3 on CursorBench and 73.7 on SWE-bench Multilingual. The model was improved via a first continued pretraining run feeding into reinforcement learning, trained across 3–4 clusters worldwide by a ~40-person team. OpenAI acquired Astral, the team behind Python tools uv, ruff, and ty, strengthening its developer platform. Anthropic expanded Claude Code with messaging app channels for persistent developer workflows. The focus in AI agents is shifting from single agents to managed fleets and runtimes, with LangChain launching LangSmith Fleet for enterprise agent management emphasizing agent identity, credential management, and auditability. Other launches include Cognition's teams of Devins, AgentUI by lvwerra, and discussions on agent runtimes with features like checkpointing and rollback. Security and permissions are emerging as critical constraints in agent system design.
not much happened today
kimi-linear-48b codex gpt-5.4 claude-code moonshot openai assemblyai langchain attention-mechanisms model-architecture inference-speed agent-feedback agent-skills multi-agent-systems knowledge-transfer cli-tools coding-agents model-deployment kimi_moonshot elonmusk yuchenj_uw nathancgy4 eliebakouch tokenbender behrouz_ali cloneofsimo fidjissimo sama gdb andrewyng itsafiz simplifyinai
Moonshot's Attention Residuals paper introduced an input-dependent attention mechanism over prior layers with a 1.25x compute advantage and less than 2% inference latency overhead, validated on Kimi Linear 48B total / 3B active. The paper sparked debate on novelty versus prior art like DeepCrossAttention and Google’s earlier work, highlighting tensions in idea novelty, citation quality, and frontier-scale validation. OpenAI's Codex showed strong momentum with over 2M weekly active users, nearly 4x growth YTD, and GPT-5.4 hitting 5T tokens/day and a $1B annualized run-rate. Codex added subagents supporting multi-agent coding workflows. Infrastructure for coding agents matured with tools like Context Hub / chub supporting agent feedback loops, AssemblyAI's skill for Claude Code and Codex, and automated skill extraction from GitHub repos yielding 40% knowledge-transfer gains. LangChain launched LangGraph CLI and open-sourced Deep Agents, recreating top coding agent workflows with planning, filesystem ops, shell access, and sub-agents.
GPT 5.4: SOTA Knowledge Work -and- Coding -and- CUA Model, OpenAI is so very back
gpt-5.4 gpt-5.4-pro openai cursor_ai perplexity_ai arena native-computer-use long-context efficiency steering benchmarking gpu-kernels attention-mechanisms algorithmic-optimization pipeline-optimization sama reach_vb scaling01 danshipper yuchenj_uw
OpenAI launched GPT-5.4 and GPT-5.4 Pro with unified mainline and Codex models, featuring native computer use, up to ~1M token context, and efficiency improvements including a new Codex
/fast mode. Benchmarks showed strong results like OSWorld-Verified 75.0% surpassing human baseline and GDPval 83% against industry pros. User feedback highlighted coding utility but raised concerns about pricing and overthinking. Integration with devtools like Cursor, Perplexity, and Arena was announced. In systems research, FlashAttention-4 (FA4) was introduced with near-matmul speed attention on Blackwell GPUs, featuring innovations like polynomial exp emulation and online softmax. "Steering mid-response" and "fewer tokens, faster speed" were emphasized as UX and efficiency improvements. not much happened today
qwen-3.5-0.8b qwen-3.5-2b qwen-3.5-4b qwen-3.5-9b codex-5.3 claude-3 alibaba ollama lm-studio openai anthropic multimodality reinforcement-learning long-context hybrid-attention on-device-ai model-deployment agent-reliability agent-observability coding-agents benchmarking runtime-optimization token-efficiency nrehiew_ kimmonismus lioronai danielhanchen theo htihle teortaxestex theprimeagen yuchenj_uw _lewtun saen_dev _philschmid omarsar0
Alibaba released the Qwen 3.5 series with models ranging from 0.8B to 9B parameters, featuring native multimodality, scaled reinforcement learning, and targeting edge and lightweight agent deployments. The models support very long context windows up to 262K tokens (extendable to 1M) and use a novel Gated DeltaNet hybrid attention architecture combining linear and full attention layers. Deployment examples include Ollama and LM Studio, with a notable 6-bit on-device demo on iPhone 17 Pro. Evaluators are cautioned that reasoning is disabled by default on smaller models. In coding agents, Codex 5.3 shows promising benchmark results on WeirdML with 79.3% accuracy, though availability and downtime remain critical challenges, especially highlighted by Claude outages. Agent reliability and observability are emphasized as cross-functional problems requiring clear success criteria and practical evaluation strategies. Studies show that using AGENTS.md and SKILL.md guardrails can significantly reduce runtime and token usage by mitigating worst-case thrashing in coding workflows.
not much happened today
gemini-3.1-pro gpt-5.2 opus-4.6 sonnet-4.6 claude-opus-4.6 google-deepmind anthropic context-arena artificial-analysis epoch-ai scaling01 retrieval benchmarking evaluation-methodology token-limits cost-efficiency instruction-following software-reasoning model-reliability dillonuzar artificialanlys yuchenj_uw theo minimax_ai epochairesearch paul_cal scaling01 metr_evals idavidrein xlr8harder htihle arena
Gemini 3.1 Pro demonstrates strong retrieval capabilities and cost efficiency compared to GPT-5.2 and Opus 4.6, though users report tooling and UI issues. The SWE-bench Verified evaluation methodology is under scrutiny for consistency, with updates bringing results closer to developer claims. Benchmarking debates arise over what frontier models truly measure, especially with ARC-AGI puzzles. Claude Opus 4.6 shows a noisy but notable 14.5-hour time horizon on software tasks, with token limits causing practical failures. Sonnet 4.6 improves significantly in code and instruction-following benchmarks, but user backlash grows due to product regressions.
OpenAI Codex App: death of the VSCode fork, multitasking worktrees, Skills Automations
codex openai agent-based-systems parallel-processing software-testing developer-workflows automation product-feedback-loop neurosymbolic-ai benchmarking sama reach_vb gdb skirano embirico ajambrosino thsottiaux nbaschez yuchenj_uw badlogicgames random_walker
OpenAI launched the Codex app on macOS as a dedicated agent-native command center for coding, featuring multiple agents in parallel, built-in worktrees for conflict isolation, skills for reusable bundles, and scheduled automations. The app emphasizes developer workflows like Plan mode for upfront task decomposition and is gaining positive adoption signals from insiders including @sama. There is movement towards ecosystem standardization of skills folders, signaling early conventions in agent tooling. Codex also exemplifies a "self-improving" product feedback loop combining humans and agents. In coding agents practice, best practices include a "test-first" approach to bug fixes, the "conductor" model where one developer manages 5-10 agents in parallel, and a neurosymbolic framing explaining why coding agents succeed due to software's verifiability and symbolic tooling. Benchmark skepticism remains about productivity studies that do not reflect agentic workflows.
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%+.
Open Responses: explicit spec for OpenAI's Responses API supported by OpenRouter, Ollama, Huggingface, vLLM, et al
gpt-5.2 opus-4.5 openai ollama vllm openrouter anthropic google-deepmind langchain llamaindex interoperable-apis agent-architecture filesystem-memory api-standardization multi-agent-systems prompt-engineering model-comparison virtual-filesystems open-source agent-ux reach_vb simonw yuchenj_uw omarsar0 jerryjliu0 hwchase17 swyx
OpenAI launched the Open Responses API spec, an open-source, multi-provider standard for interoperable LLM APIs designed to simplify agent stacks and tooling. Early adopters like ollama and vLLM support the spec, while notable absences include anthropic and google-deepmind. Agent design insights from Cursor emphasize explicit roles and planning over mega-agent models, with GPT-5.2 outperforming Opus 4.5 in long runs. The emerging dominant context/memory abstraction for agents is a filesystem-as-memory approach, championed by llamaindex and langchain, using virtual filesystems often backed by databases like Postgres. LangChain also shipped an open-source desktop interface for agent orchestration called openwork. This news highlights advances in API standardization, agent architecture, and memory abstractions in AI development.
not much happened today
claude-max anthropic openai ai21-labs github cline model-agnostic model-context-protocol tooling skills concurrency transactional-workspaces context-engineering file-centric-workspaces rate-limiting agent-workspaces yuchenj_uw andersonbcdefg gneubig matan_sf scaling01 reach_vb _philschmid claude_code code jamesmontemagno cline danstripper omarsar0
Anthropic tightens usage policies for Claude Max in third-party apps, prompting builders to adopt model-agnostic orchestration and BYO-key defaults to mitigate platform risks. The Model Context Protocol (MCP) is evolving into a key tooling plane with OpenAI MCP Server and mcp-cli enhancing tool discovery and token efficiency. The concept of skills as modular, versioned behaviors gains traction, with implementations in Claude Code, GitHub Copilot, and Cline adding websearch tooling. AI21 Labs addresses concurrency challenges in agent workspaces using git worktrees for transactional parallel writes, while long-horizon agents focus on context engineering and persistent file-centric workspaces.
xAI raises $20B Series E at ~$230B valuation
grok-5 claude-code xai nvidia cisco fidelity valor-equity-partners qatar-investment-authority mgx stepstone-group baron-capital-group hugging-face amd ai-infrastructure supercomputing robotics ai-hardware agentic-ai context-management token-optimization local-ai-assistants aakash_gupta fei-fei_li lisa_su clementdelangue thom_wolf saradu omarsar0 yuchenj_uw _catwu cursor_ai
xAI, Elon Musk's AI company, completed a massive $20 billion Series E funding round, valuing it at about $230 billion with investors like Nvidia, Cisco Investments, and others. The funds will support AI infrastructure expansion including Colossus I and II supercomputers and training Grok 5, leveraging data from X's 600 million monthly active users. At CES 2026, the focus was on "AI everywhere" with a strong emphasis on AI-first hardware and integration between NVIDIA and Hugging Face's LeRobot for robotics development. The Reachy Mini robot is gaining traction as a consumer robotics platform. In software, Claude Code is emerging as a popular local/private coding assistant, with new UI features in Claude Desktop and innovations like Cursor's dynamic context reducing token usage by nearly 47% in multi-MCP setups. "The 600 million MAU figure in xAI’s announcement combines X platform users with Grok users. That’s a clever framing choice."
not much happened today
glm-4.7 claude-code z.ai meta-ai-fair manus replit agentic-architecture context-engineering application-layer code-generation agent-habitats ai-native-llm ipo inference-infrastructure programming-paradigms zixuanli_ jietang yuchenj_uw sainingxie amasad hidecloud imjaredz random_walker
Z.ai (GLM family) IPO in Hong Kong on Jan 8, 2026, aiming to raise $560M at HK$4.35B, marking it as the "first AI-native LLM company" public listing. The IPO highlights GLM-4.7 as a starting point. Meta AI acquired Manus for approximately $4–5B, with Manus achieving $100M ARR in 8–9 months, illustrating the value of application-layer differentiation over proprietary models. Manus focuses on agentic architecture, context engineering, and general primitives like code execution and browser control, emphasizing "agent habitats" as a competitive moat. Discussions around Claude Code highlight skepticism about "vibe coding," advocating for disciplined, framework-like AI-assisted programming practices.
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.
not much happened today
trillium gemini-2.5-pro gemini-deepthink google huawei epoch-ai deutsche-telekom nvidia anthropic reka-ai weaviate deepmind energy-efficiency datacenters mcp context-engineering instruction-following embedding-models math-reasoning benchmarking code-execution sundarpichai yuchenj_uw teortaxestex epochairesearch scaling01 _avichawla rekaailabs anthropicai douwekiela omarsar0 nityeshaga goodside iscienceluvr lmthang
Google's Project Suncatcher prototypes scalable ML compute systems in orbit using solar energy with Trillium-generation TPUs surviving radiation, aiming for prototype satellites by 2027. China's 50% electricity subsidies for datacenters may offset chip efficiency gaps, with Huawei planning gigawatt-scale SuperPoDs for DeepSeek by 2027. Epoch launched an open data center tracking hub, and Deutsche Telekom and NVIDIA announced a $1.1B Munich facility with 10k GPUs. In agent stacks, MCP (Model-Compute-Platform) tools gain traction with implementations like LitServe, Claude Desktop, and Reka's MCP server for VS Code. Anthropic emphasizes efficient code execution with MCP. Context engineering shifts focus from prompt writing to model input prioritization, with reports and tools from Weaviate, Anthropic, and practitioners highlighting instruction-following rerankers and embedding approaches. DeepMind's IMO-Bench math reasoning suite shows Gemini DeepThink achieving high scores, with a ProofAutoGrader correlating strongly with human grading. Benchmarks and governance updates include new tasks and eval sharing in lighteval.
not much happened today
vllm chatgpt-atlas langchain meta microsoft openai pytorch ray claude agent-frameworks reinforcement-learning distributed-computing inference-correctness serving-infrastructure browser-agents security middleware runtime-systems documentation hwchase17 soumithchintala masondrxy robertnishihara cryps1s yuchenj_uw
LangChain & LangGraph 1.0 released with major updates for reliable, controllable agents and unified docs, emphasizing "Agent Engineering." Meta introduced PyTorch Monarch and TorchForge for distributed programming and reinforcement learning, enabling large-scale agentic systems. Microsoft Learn MCP server now integrates with tools like Claude Code and VS Code for instant doc querying, accelerating grounded agent workflows. vLLM improved inference correctness with token ID returns and batch-invariant inference, collaborating with Ray for orchestration in PyTorch Foundation. OpenAI launched ChatGPT Atlas, a browser agent with contextual Q&A and advanced safety features, though early users note maturity challenges and caution around credential access.
ChatGPT Atlas: OpenAI's AI Browser
gemini atlas openai google langchain ivp capitalg sapphire sequoia benchmark agent-mode browser-memory chromium finetuning moe lora agent-runtime observability software-development funding kevinweil bengoodger fidjissimo omarsar0 yuchenj_uw nickaturley raizamrtn hwchase17 bromann casper_hansen_ corbtt
OpenAI launched the Chromium fork AI browser Atlas for macOS, featuring integrated Agent mode and browser memory with local login capabilities, aiming to surpass Google's Gemini in Chrome. The launch received mixed reactions regarding reliability and privacy. LangChain raised a $125M Series B at a $1.25B valuation, releasing v1.0 agent engineering stack with significant adoption including 85M+ OSS downloads/month and usage by ~35% of the Fortune 500. The ecosystem also saw updates like vLLM's MoE LoRA expert finetuning support.
GDPVal finding: Claude Opus 4.1 within 95% of AGI (human experts in top 44 white collar jobs)
claude-4.1-opus gpt-5-high gptnext gemini-2.5-flash gemini-2.5-flash-lite deepseek-v3.1-terminus google-chirp-2 qwen-2.5b openai anthropic google nvidia artificial-analysis deepseek benchmarking agentic-ai tool-use long-context speech-to-text model-evaluation reasoning pricing model-performance kevinweil gdb dejavucoder yuchenj_uw lhsummers
OpenAI's Evals team released GDPval, a comprehensive evaluation benchmark covering 1,320 tasks across 44 predominantly digital occupations, assessing AI models against human experts with 14 years average experience. Early results show Claude 4.1 Opus outperforming human experts in most categories and GPT-5 high trailing behind, with projections that GPTnext could match human performance by mid-2026. The benchmark is positioned as a key metric for policymakers and labor impact forecasting. Additionally, Artificial Analysis reported improvements in Gemini 2.5 Flash/Flash-Lite and DeepSeek V3.1 Terminus models, alongside new speech-to-text benchmarks (AA-WER) highlighting leaders like Google Chirp 2 and NVIDIA Canary Qwen2.5B. Agentic AI advances include Kimi OK Computer, an OS-like agent with extended tool capabilities and new vendor verification tools.
Qwen3-Next-80B-A3B-Base: Towards Ultimate Training & Inference Efficiency
qwen3-next qwen3 mixtral-8x7b gemini-2.5-pro alibaba mistral-ai deepseek snowflake hugging-face baseten nvidia mixture-of-experts model-sparsity gated-attention hybrid-architecture rmsnorm model-stability model-training inference-optimization multi-token-prediction model-deployment justinlin610 teortaxestex yuchenj_uw
MoE (Mixture of Experts) models have become essential in frontier AI models, with Qwen3-Next pushing sparsity further by activating only 3.7% of parameters (3B out of 80B) using a hybrid architecture combining Gated DeltaNet and Gated Attention. This new design includes 512 total experts (10 routed + 1 shared), Zero-Centered RMSNorm for stability, and improved MoE router initialization, resulting in ~10× cheaper training and 10× faster inference compared to previous models. Alibaba's Qwen3-Next reportedly outperforms Gemini-2.5-Flash-Thinking and approaches the flagship 235B model's performance, with deployments on Hugging Face, Baseten, and native vLLM support for efficient inference.
not much happened today
glm-4.5 glm-4.5-air qwen3-coder qwen3-235b kimi-k2 grok-imagine wan-2.2 smollm3 figure-01 figure-02 vitpose++ chatgpt zhipu-ai alibaba moonshot-ai x-ai figure openai runway mlx ollama deeplearningai model-releases model-performance moe image-generation video-generation pose-estimation robotics training-code-release interactive-learning in-context-learning yuchenj_uw corbtt reach_vb ollama deeplearningai gdb sama c_valenzuelab adcock_brett skalskip92 loubnabenallal1 hojonathanho ostrisai
Chinese AI labs have released powerful open-source models like GLM-4.5 and GLM-4.5-Air from Zhipu AI, Qwen3 Coder and Qwen3-235B from Alibaba, and Kimi K2 from Moonshot AI, highlighting a surge in permissively licensed models. Zhipu AI's GLM-4.5 is a 355B parameter MoE model competitive with Claude 4 Opus and Gemini 2.5 Pro. Alibaba's Qwen3 Coder shows strong code generation performance with a low edit failure rate, while Moonshot AI's Kimi K2 is a 1 trillion-parameter MoE model surpassing benchmarks like LiveCodeBench. In video and image generation, xAI launched Grok Imagine, and Wan2.2 impressed with innovative image-to-video generation. Robotics advances include Figure's Figure-01 and Figure-02 humanoid robots and ViTPose++ for pose estimation in basketball analysis. SmolLM3 training and evaluation code was fully released under Apache 2.0. OpenAI introduced Study Mode in ChatGPT to enhance interactive learning, and Runway rolled out Runway Aleph, a new in-context video model for multi-task visual generation. The community notes a competitive disadvantage for organizations avoiding these Chinese open-source models. "Orgs avoiding these models are at a significant competitive disadvantage," noted by @corbtt.
not much happened today
glm-4.5 glm-4.5-air qwen3-coder qwen3-235b kimi-k2 wan-2.2 grok-imagine smollm3 figure-01 figure-02 vitpose++ zhipu-ai alibaba moonshot-ai x-ai ideogram figure smollm openai model-releases moe model-benchmarking image-generation video-generation pose-estimation robotics training-code-release apache-license yuchenj_uw corbtt cline reach_vb ollama deeplearningai ostrisai hojonathanho adcock_brett skalskip92 loubnabenallal1
Chinese labs have released a wave of powerful, permissively licensed models in July, including Zhipu AI's GLM-4.5 and GLM-4.5-Air, Alibaba's Qwen3 Coder and Qwen3-235B, and Moonshot AI's Kimi K2. These models feature large-scale Mixture of Experts architectures with active parameters ranging from 3B to 32B and context windows up to 256K tokens. Zhipu AI's GLM-4.5 competes with Claude 4 Opus and Gemini 2.5 Pro in benchmarks. Moonshot AI's Kimi K2 is a 1 trillion-parameter MoE model surpassing other open-weight models on LiveCodeBench and AceBench. In video and image generation, xAI launched Grok Imagine, and Wan2.2 impressed with its Image-to-Video approach. Ideogram released a character consistency model. Robotics advances include Figure's Figure-01 and Figure-02 humanoid robots and ViTPose++ for pose estimation in basketball analysis. The SmolLM3 training and evaluation code was fully released under an Apache 2.0 license. "Orgs avoiding these Chinese open-source models are at a significant competitive disadvantage," noted by @corbtt.
not much happened today
qwen3-coder-480b-a35b-instruct kimi-k2 alibaba openrouterai togethercompute vllm_project unslothai white-house code-generation benchmarking model-integration context-windows open-source national-security infrastructure ai-policy fchollet clementdelangue scaling01 aravsrinivas rasbt gregkamradt yuchenj_uw
Alibaba announced the release of Qwen3-Coder-480B-A35B-Instruct, an open agentic code model with 480B parameters and 256K context length, praised for rapid development and strong coding performance. Benchmark claims of 41.8% on ARC-AGI-1 faced skepticism from Fran ois Chollet and others due to reproducibility issues. The model quickly integrated into ecosystems like vLLM, Dynamic GGUFs, and OpenRouterAI. The White House unveiled a new AI Action Plan emphasizing Innovation, Infrastructure, and International Diplomacy, linking AI leadership to national security and prioritizing compute access for the Department of Defense. The plan sparked debate on open vs. closed-source AI, with calls from Clement Delangue to embrace open science to maintain US AI competitiveness.
not much happened today
kimi-k2 grok-4 gpt-5 gemini-2.5 gemini-embedding cognition windsurf moonshot-ai x-ai openai google stanfordnlp huggingface mixture-of-experts model-training model-performance fine-tuning benchmarking agentic-ai model-bugs embedding-models sama hardmaru jeremyphoward akhaliq teortaxestex yuchenj_uw demishassabis
Cognition is acquiring the remaining assets of Windsurf after a significant weekend deal. Moonshot AI released Kimi K2, an open-source, MIT-licensed agentic model with 1 Trillion total / 32B active parameters using a Mixture-of-Experts architecture, trained on 15.5 Trillion tokens with the MuonClip optimizer, showing top performance on benchmarks like EQ-Bench and Creative Writing. xAI launched Grok-4, ranking 5th on IQ Bench but with notable quirks including a bug causing it to respond only with "Heavy" and a high frequency of Elon Musk mentions. Rumors about OpenAI delaying an open-source model release surfaced, with speculation about CEO sama's PR strategy and a possible GPT-5 launch in September. The Gemini 2.5 paper was released with 3,295 authors, and Google introduced its Gemini Embedding model, topping the MTEB leaderboard.
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.
Grok 4: xAI succeeds in going from 0 to new SOTA LLM in 2 years
grok-4 grok-4-heavy claude-4-opus xai perplexity-ai langchain cursor cline model-releases benchmarking long-context model-pricing model-integration voice performance scaling gpu-optimization elonmusk aravsrinivas igor_babuschkin yuchenj_uw
xAI launched Grok 4 and Grok 4 Heavy, large language models rumored to have 2.4 trillion parameters and trained with 100x more compute than Grok 2 on 100k H100 GPUs. Grok 4 achieved new state-of-the-art results on benchmarks like ARC-AGI-2 (15.9%), HLE (50.7%), and Vending-Bench, outperforming models such as Claude 4 Opus. The model supports a 256K context window and is priced at $3.00/M input tokens and $15.00/M output tokens. It is integrated into platforms like Cursor, Cline, LangChain, and Perplexity Pro/Max. The launch was accompanied by a controversial voice mode and sparked industry discussion about xAI's rapid development pace, with endorsements from figures like Elon Musk and Arav Srinivas.
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.
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.
Granola launches team notes, while Notion launches meeting transcription
gpt-4.1 gpt-4o-mini gpt-4.1-mini claude-opus claude-sonnet claude-o3 qwen3 seed1.5-vl llama-4 am-thinking-v1 openai anthropic alibaba meta-ai-fair huggingface granola coding instruction-following benchmarking model-releases reasoning image-generation collaborative-software model-performance kevinweil scaling01 steph_palazzolo andersonbcdefg reach_vb yuchenj_uw qtnx_ _akhaliq risingsayak
GPT-4.1 is now available in ChatGPT for Plus, Pro, and Team users, focusing on coding and instruction following, with GPT 4.1 mini replacing GPT 4o mini. Anthropic is releasing new Claude models including Claude Opus and Claude Sonnet, though some criticism about hallucinations in Claude O3 was noted. Alibaba shared the Qwen3 Technical Report with strong benchmark results from Seed1.5-VL. Meta FAIR announced new models and datasets but faced criticism on Llama 4. AM-Thinking-v1 launched on Hugging Face as a 32B scale reasoning model. Granola raised $43M in Series B and launched Granola 2.0 with a Notion-like UI. The AI ecosystem shows rapid iteration and cloning of ideas, emphasizing execution and distribution.
Google's Agent2Agent Protocol (A2A)
kimi-vl-a3b gpt-4o llama-4-scout llama-4-maverick llama-4-behemoth deepcoder-14b o3-mini o1 llama-3.1-nemotron-ultra-253b deepseek-r1 google google-deepmind moonshot-ai meta-ai-fair uc-berkeley openai nvidia hugging-face togethercompute deepseek agent-interoperability multimodality vision math reinforcement-learning coding model-training open-source model-benchmarking context-windows streaming push-notifications enterprise-authentication model-release reach_vb _akhaliq epochairesearch artificialanlys winglian danielhanchen yuchenj_uw jeremyphoward
Google Cloud Next announcements featured the launch of Google and DeepMind's full MCP support and a new Agent to Agent protocol designed for agent interoperability with multiple partners. The protocol includes components like the Agent Card, Task communication channels, Enterprise Auth and Observability, and Streaming and Push Notification support. On the model front, Moonshot AI released Kimi-VL-A3B, a multimodal model with 128K context and strong vision and math benchmark performance, outperforming gpt-4o. Meta AI introduced smaller versions of llama-4 family models: llama-4-scout and llama-4-maverick, with a larger Behemoth model still in training. DeepCoder 14B from UC Berkeley is an open-source coding model rivaling openai's o3-mini and o1 models, trained with reinforcement learning on 24K coding problems. Nvidia released llama-3.1-nemotron-ultra-253b on Hugging Face, noted for beating llama-4-behemoth and maverick and competing with deepseek-r1.
DeepCoder: A Fully Open-Source 14B Coder at O3-mini Level
deepcoder-14b o3-mini o1 gemini-2.5-pro kimi-vl-a3b gpt-4o llama-4-scout maverick behemoth gen-4-turbo imagen-3 together-ai agentica opena bytedance google-deepmind moonshot-ai meta-ai-fair runway open-source reinforcement-learning code-generation multimodality model-training mixture-of-experts l2-normalization image-generation model-performance context-windows philschmid lepikhin reach_vb akhaliq yuchenj_uw epochairesearch danielhanchen c_valenzuelab
Together AI and Agentica released DeepCoder-14B, an open-source 14B parameter coding model rivaling OpenAI's o3-mini and o1 on coding benchmarks, trained with an open-source RL framework from ByteDance and costing about $26,880. Google DeepMind launched Gemini 2.5 Pro with experimental "Flash" versions available to subscribers. Moonshot AI introduced Kimi-VL-A3B, a multimodal model with 128K context outperforming gpt-4o on vision and math benchmarks. Meta AI released Llama 4 Scout and Maverick, with a larger Behemoth model in training, featuring mixture-of-experts and L2 norm techniques. Runway launched Gen-4 Turbo with 10x better results than Gen-3 at the same cost. Google announced Imagen 3, a high-quality text-to-image model now in Vertex AI, enabling easier object removal. The report highlights open-source contributions, reinforcement learning training optimizations, and significant model performance improvements across coding, multimodal, and image generation domains.
Llama 4's Controversial Weekend Release
llama-4 llama-3 llama-3-2 meta mixture-of-experts early-fusion attention-mechanisms fp8-training training-data benchmarking model-performance model-release multimodality open-models ahmad_al_dahle ylecun reach_vb yuchenj_uw
Meta released Llama 4, featuring two new medium-size MoE open models and a promised 2 Trillion parameter "behemoth" model, aiming to be the largest open model ever. The release included advanced training techniques like Chameleon-like early fusion with MetaCLIP, interleaved chunked attention without RoPE, native FP8 training, and training on up to 40 trillion tokens. Despite the hype, the release faced criticism for lack of transparency compared to Llama 3, implementation issues, and poor performance on some benchmarks. Meta leadership, including Ahmad Al Dahle, denied allegations of training on test sets. The smallest Scout model at 109B parameters is too large for consumer GPUs, and the claimed 10 million token context is disputed. The community response has been mixed, with some praising the openness and others pointing out discrepancies and quality concerns.
not much happened today
gpt-4.5 gpt-4 gpt-4o o1 claude-3.5-sonnet claude-3.7 claude-3-opus deepseek-v3 grok-3 openai anthropic perplexity-ai deepseek scaling01 model-performance humor emotional-intelligence model-comparison pricing context-windows model-size user-experience andrej-karpathy jeremyphoward abacaj stevenheidel yuchenj_uw aravsrinivas dylan522p random_walker
GPT-4.5 sparked mixed reactions on Twitter, with @karpathy noting users preferred GPT-4 in a poll despite his personal favor for GPT-4.5's creativity and humor. Critics like @abacaj highlighted GPT-4.5's slowness and questioned its practical value and pricing compared to other models. Performance-wise, GPT-4.5 ranks above GPT-4o but below o1 and Claude 3.5 Sonnet, with Claude 3.7 outperforming it on many tasks yet GPT-4.5 praised for its humor and "vibes." Speculation about GPT-4.5's size suggests around 5 trillion parameters. Discussions also touched on pricing disparities, with Perplexity Deep Research at $20/month versus ChatGPT at $200/month. The emotional intelligence and humor of models like Claude 3.7 were also noted.
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.
$1150m for SSI, Sakana, You.com + Claude 500m context
olmo llama2-13b-chat claude claude-3.5-sonnet safe-superintelligence sakana-ai you-com perplexity-ai anthropic ai2 mixture-of-experts model-architecture model-training gpu-costs retrieval-augmented-generation video-generation ai-alignment enterprise-ai agentic-ai command-and-control ilya-sutskever mervenoyann yuchenj_uw rohanpaul_ai ctojunior omarsar0
Safe Superintelligence raised $1 billion at a $5 billion valuation, focusing on safety and search approaches as hinted by Ilya Sutskever. Sakana AI secured a $100 million Series A funding round, emphasizing nature-inspired collective intelligence. You.com pivoted to a ChatGPT-like productivity agent after a $50 million Series B round, while Perplexity AI raised over $250 million this summer. Anthropic launched Claude for Enterprise with a 500 million token context window. AI2 released a 64-expert Mixture-of-Experts (MoE) model called OLMo, outperforming Llama2-13B-Chat. Key AI research trends include efficient MoE architectures, challenges in AI alignment and GPU costs, and emerging AI agents for autonomous tasks. Innovations in AI development feature command and control for video generation, Retrieval-Augmented Generation (RAG) efficiency, and GitHub integration under Anthropic's Enterprise plan. "Our logo is meant to invoke the idea of a school of fish coming together and forming a coherent entity from simple rules as we want to make use of ideas from nature such as evolution and collective intelligence in our research."