All tags
Person: "_jasonwei"
not much happened today
kimi-k2 gpt-4.1 voxtral goedel-prover-v2 llama-3 mistral-ai moonshot-ai nous-research google-deepmind openai groq anthropic speech-recognition mixture-of-experts benchmarking dataset-release model-architecture theorem-proving reinforcement-learning asymmetry-of-verification inference-speed model-performance cline _jasonwei
Mistral released Voxtral, claimed as the world's best open speech recognition models, available via API and Hugging Face. Moonshot AI launched Kimi K2, a trillion-parameter Mixture-of-Experts (MoE) model, outperforming GPT-4.1 on benchmarks with 65.4% on SWE-Bench Verified and achieving 200 tokens/second inference speed on Groq hardware. Nous Research open-sourced the Hermes 3 dataset with 1 million samples, aiding SOTA models on the Llama-3 series. Google DeepMind introduced the Mixture-of-Recursions (MoR) architecture promising 2x inference speed and 50% parameter reduction but faced skepticism. Goedel-Prover V2 topped the PutnamBench theorem proving benchmark. AtCoder World Finals saw a human winner with OpenAI placing second. Research highlights include Jason Wei's insights on reinforcement learning and the "Verifier's Law" emphasizing the asymmetry of verification in AI training.
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.
The AI Search Wars Have Begun — SearchGPT, Gemini Grounding, and more
gpt-4o o1-preview claude-3.5-sonnet universal-2 openai google gemini nyt perplexity-ai glean nvidia langchain langgraph weights-biases cohere weaviate fine-tuning synthetic-data distillation hallucinations benchmarking speech-to-text robotics neural-networks ai-agents sam-altman alexalbert__ _jasonwei svpino drjimfan virattt
ChatGPT launched its search functionality across all platforms using a fine-tuned version of GPT-4o with synthetic data generation and distillation from o1-preview. This feature includes a Chrome extension promoted by Sam Altman but has issues with hallucinations. The launch coincides with Gemini introducing Search Grounding after delays. Notably, The New York Times is not a partner due to a lawsuit against OpenAI. The AI search competition intensifies with consumer and B2B players like Perplexity and Glean. Additionally, Claude 3.5 Sonnet achieved a new benchmark record on SWE-bench Verified, and a new hallucination evaluation benchmark, SimpleQA, was introduced. Other highlights include the Universal-2 speech-to-text model with 660M parameters and HOVER, a neural whole-body controller for humanoid robots trained in NVIDIA Isaac simulation. AI hedge fund teams using LangChain and LangGraph were also showcased. The news is sponsored by the RAG++ course featuring experts from Weights & Biases, Cohere, and Weaviate.
Gemini Live
gemini-1.5-pro genie falcon-mamba gemini-1.5 llamaindex google anthropic tii supabase perplexity-ai llamaindex openai hugging-face multimodality benchmarking long-context retrieval-augmented-generation open-source model-releases model-integration model-performance software-engineering linear-algebra hugging-face-hub debugging omarsar0 osanseviero dbrxmosaicai alphasignalai perplexity_ai _jasonwei svpino
Google launched Gemini Live on Android for Gemini Advanced subscribers during the Pixel 9 event, featuring integrations with Google Workspace apps and other Google services. The rollout began on 8/12/2024, with iOS support planned. Anthropic released Genie, an AI software engineering system achieving a 57% improvement on SWE-Bench. TII introduced Falcon Mamba, a 7B attention-free open-access model scalable to long sequences. Benchmarking showed that longer context lengths do not always improve Retrieval-Augmented Generation. Supabase launched an AI-powered Postgres service dubbed the "ChatGPT of databases," fully open source. Perplexity AI partnered with Polymarket to integrate real-time probability predictions into search results. A tutorial demonstrated a multimodal recipe recommender using Qdrant, LlamaIndex, and Gemini. An OpenAI engineer shared success tips emphasizing debugging and hard work. The connection between matrices and graphs in linear algebra was highlighted for insights into nonnegative matrices and strongly connected components. Keras 3.5.0 was released with Hugging Face Hub integration for model saving and loading.