All tags
Company: "danielhanchen"
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
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%+.
not much happened today
nomos-1 axiomprover devstral-2-small deepseek-v3.2 claude-code cursor-2.2 claude-opus-4.5 gpt-5 claude-sonnet-4.5 gemini-3-pro llama qwen mistral gemma nousresearch thinkymachines mistral-ai deepseek anthropic cursor microsoft langchain-ai openai gemini intel vllm_project danielhanchen math formal-reasoning agentic-systems asynchronous-execution multi-agent-systems observability benchmarking quantization post-training-quantization training-speedup kernel-optimization inference-efficiency
NousResearch's Nomos 1 is a 30B open math model achieving a top Putnam score with only ~3B active parameters, enabling consumer Mac inference. AxiomProver also posts top Putnam results using ThinkyMachines' RL stack. Mistral's Devstral 2 Small outperforms DeepSeek v3.2 in 71% of preferences with better speed and cost. Anthropic's Claude Code introduces asynchronous agent execution. Cursor 2.2 adds deep agent primitives like Debug and Plan Modes. VS Code launches unified agent chat sessions improving multi-agent workflows. LangChain releases "Polly" for agent observability. The Stirrup harness leads OpenAI GDPval benchmarks with Claude Opus 4.5, GPT-5, and Gemini 3 Pro following. Advances in quantization include vLLM integrating Intel's AutoRound PTQ for efficient serving. Unsloth achieves up to 3× training speedups with new kernels across Llama, Qwen, Mistral, and Gemma models. "Compositional reasoning + specialized post-training under constrained active params can rival frontier closed models on formal math."
not much happened today
gemini-1.5-pro claude-3 chatgpt langchain meta-ai-fair hugging-face openrouter google-ai microsoft openai anthropic agent-ops observability multi-turn-evaluation reinforcement-learning distributed-training api model-stability user-intent-clustering software-development project-management code-generation hwchase17 ankush_gola11 whinthorn koylanai _lewtun bhutanisanyam1 thom_wolf danielhanchen cline canvrno pashmerepat mustafasuleyman yusuf_i_mehdi jordirib1 fidjissimo bradlightcap mikeyk alexalbert__
LangSmith launched the Insights Agent with multi-turn evaluation for agent ops and observability, improving failure detection and user intent clustering. Meta PyTorch and Hugging Face introduced OpenEnv, a Gymnasium-style API and hub for reproducible agentic environments supporting distributed training. Discussions highlighted the importance of provider fidelity in agent coding, with OpenRouter's exacto filter improving stability. Builder UX updates include Google AI Studio's Annotation mode for Gemini code changes, Microsoft's Copilot Mode enhancements in Edge, and OpenAI's Shared Projects and Company Knowledge features for ChatGPT Business. Claude added project-scoped Memory. In reinforcement learning, Meta's ScaleRL proposes a methodology to predict RL scaling outcomes for LLMs with improved efficiency and stability.
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.
not much happened today
gemma-3n glm-4.1v-thinking deepseek-r1t2 mini-max-m1 o3 claude-4-opus claude-sonnet moe-72b meta scale-ai unslothai zhipu-ai deepseek huawei minimax-ai allenai sakana-ai-labs openai model-performance vision conv2d float16 training-loss open-source model-benchmarks moe load-balancing scientific-literature-evaluation code-generation adaptive-tree-search synthesis-benchmarks alexandr_wang natfriedman steph_palazzolo thegregyang teortaxes_tex denny_zhou agihippo danielhanchen osanseviero reach_vb scaling01 ndea
Meta has hired Scale AI CEO Alexandr Wang as its new Chief AI Officer, acquiring a 49% non-voting stake in Scale AI for $14.3 billion, doubling its valuation to ~$28 billion. This move is part of a major talent shuffle involving Meta, OpenAI, and Scale AI. Discussions include the impact on Yann LeCun's influence at Meta and potential responses from OpenAI. In model news, Gemma 3N faces technical issues like vision NaNs and FP16 overflows, with fixes from UnslothAI. Chinese open-source models like GLM-4.1V-Thinking by Zhipu AI and DeepSeek R1T2 show strong performance and speed improvements. Huawei open-sourced a 72B MoE model with a novel load balancing solution. The MiniMax-M1 hybrid MoE model leads math benchmarks on the Text Arena leaderboard. AllenAI launched SciArena for scientific literature evaluation, where o3 outperforms others. Research from Sakana AI Labs introduces AB-MCTS for code generation, improving synthesis benchmarks.
not much happened today
deepseek-r1-0528 o3 gemini-2.5-pro claude-opus-4 deepseek_ai openai gemini meta-ai-fair anthropic x-ai ollama hugging-face alibaba bytedance xiaomi reasoning reinforcement-learning benchmarking quantization local-inference model-evaluation open-weights transparency post-training agentic-benchmarks long-context hallucination-detection teortaxestex wenfeng danielhanchen awnihannun reach_vb abacaj
DeepSeek R1-0528 release brings major improvements in reasoning, hallucination reduction, JSON output, and function calling, matching or surpassing closed models like OpenAI o3 and Gemini 2.5 Pro on benchmarks such as Artificial Analysis Intelligence Index, LiveBench, and GPQA Diamond. The model ranks #2 globally in open weights intelligence, surpassing Meta AI, Anthropic, and xAI. Open weights and technical transparency have fueled rapid adoption across platforms like Ollama and Hugging Face. Chinese AI labs including DeepSeek, Alibaba, ByteDance, and Xiaomi now match or surpass US labs in model releases and intelligence, driven by open weights strategies. Reinforcement learning post-training is critical for intelligence gains, mirroring trends seen at OpenAI. Optimized quantization techniques (1-bit, 4-bit) and local inference enable efficient experimentation on consumer hardware. New benchmarks like LisanBench test knowledge, planning, memory, and long-context reasoning, with OpenAI o3 and Claude Opus 4 leading. Discussions highlight concerns about benchmark contamination and overemphasis on RL-tuned gains.
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.
not much happened today
gemini-2.5-pro chatgpt deepseek-v3 qwen-2.5 claude-3.5-sonnet claude-3.7-sonnet google anthropic openai llama_index langchain runway deepseek math benchmarking chains-of-thought model-performance multi-agent-systems agent-frameworks media-generation long-horizon-planning code-generation rasbt danielhanchen hkproj
Gemini 2.5 Pro shows strengths and weaknesses, notably lacking LaTex math rendering unlike ChatGPT, and scored 24.4% on the 2025 US AMO. DeepSeek V3 ranks 8th and 12th on recent leaderboards. Qwen 2.5 models have been integrated into the PocketPal app. Research from Anthropic reveals that Chains-of-Thought (CoT) reasoning is often unfaithful, especially on harder tasks, raising safety concerns. OpenAI's PaperBench benchmark shows AI agents struggle with long-horizon planning, with Claude 3.5 Sonnet achieving only 21.0% accuracy. CodeAct framework generalizes ReAct for dynamic code writing by agents. LangChain explains multi-agent handoffs in LangGraph. Runway Gen-4 marks a new phase in media creation.
Gemma 3 beats DeepSeek V3 in Elo, 2.0 Flash beats GPT4o with Native Image Gen
gemma-3 gemini-1.5-pro gemini-2 o1-preview o3-mini-high deepseek-v3 claude-3.7-sonnet qwen-2.5-max google-deepmind openai multimodality multilinguality context-window quantization image-generation model-benchmarking model-performance vision reach_vb _philschmid danielhanchen lmarena_ai osanseviero
Google DeepMind launched the Gemma 3 family of models featuring a 128k context window, multimodal input (image and video), and multilingual support for 140+ languages. The Gemma 3-27B model ranks among the top open models on LMArena benchmarks, outperforming several competitors and matching Gemini-1.5-Pro on benchmarks. Additionally, Gemini 2 introduced Flash Native Image Generation with advanced image editing capabilities, a feature teased by OpenAI but not launched. The updates highlight significant advances in context length, multimodality, and model efficiency via quantization.
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
claude-3.7-sonnet claude-3.7 deepseek-r1 o3-mini deepseek-v3 gemini-2.0-pro gpt-4o qwen2.5-coder-32b-instruct anthropic perplexity-ai amazon google-cloud deepseek_ai coding reasoning model-benchmarking agentic-workflows context-window model-performance open-source moe model-training communication-libraries fp8 nvlink rdma cli-tools skirano omarsar0 reach_vb artificialanlys terryyuezhuo _akhaliq _philschmid catherineols goodside danielhanchen
Claude 3.7 Sonnet demonstrates exceptional coding and reasoning capabilities, outperforming models like DeepSeek R1, O3-mini, and GPT-4o on benchmarks such as SciCode and LiveCodeBench. It is available on platforms including Perplexity Pro, Anthropic, Amazon Bedrock, and Google Cloud, with pricing at $3/$15 per million tokens. Key features include a 64k token thinking mode, 200k context window, and the CLI-based coding assistant Claude Code. Meanwhile, DeepSeek released DeepEP, an open-source communication library optimized for MoE model training and inference with support for NVLink, RDMA, and FP8. These updates highlight advancements in coding AI and efficient model training infrastructure.
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.