MODELS · NvidiaMODELS
Models & research
Foundation models, benchmarks and research papers.
MODELS · Nvidia
MODELS · TencentTencent launches Hy4 flash, a cheaper multimodal model with 12 billion active parameters
2 MIN READ · NOV 17, 2026
MODELS · OpenAIOpenAI embeds four outside safety assessors: Apollo Research, METR, Transluce and Goodfire
2 MIN READ · NOV 17, 2026
MODELS · InternLMShanghai AI Lab quietly publishes Sage S2.1 Flash, a 321B-parameter open scientific model
2 MIN READ · NOV 16, 2026
MODELS · OpenAIOpenAI opens a science blog with a physicist's account of GPT-6 Astra's role in a lattice Standard Model proposal
2 MIN READ · NOV 16, 2026
MODELS · BlandBland opens a research preview of Tofu-1, a full-duplex speech model built for noisy phone calls
2 MIN READ · NOV 16, 2026
MODELS · AlibabaAlibaba officially launches Qwen4-Max, pointing to security-focused training and tighter misuse safeguards
2 MIN READ · NOV 16, 2026
MODELS · BenchmarkSingularity-Bench: Claude Opus 5.5 agent matches or beats human-built baselines on six AI infrastructure tasks
2 MIN READ · NOV 15, 2026
MODELS · SalesforceSalesforce releases SOV-Transcribe-1, an open-weight speech recognition model for self-hosting
2 MIN READ · NOV 15, 2026
MODELS · TencentTencent Hunyuan launches Hy Realtime, a voice model that can call tools and sing mid-conversation
2 MIN READ · NOV 14, 2026
MODELS · BaiduBaidu opens ERNIE 5.5 Flash Preview on Qianfan at RMB 1 per million input tokens and launches a Token Exchange model marketplace
2 MIN READ · NOV 14, 2026
MODELS · MetaMeta releases Muse Ember, its smallest Muse model, priced for high-volume agent work
2 MIN READ · NOV 14, 2026
MODELS · AlibabaAlibaba opens a two-week preview of Qwen4-Max with a one-million-token context window
2 MIN READ · NOV 14, 2026
MODELS · Google DeepMindDeepMind's AlphaGenome 2 predicts how DNA variants act in specific cell types
2 MIN READ · NOV 13, 2026
MODELS · GoogleGoogle releases Gemini 4 Hydrogen, a cheaper Gemini 4 model for everyday agentic work
2 MIN READ · NOV 13, 2026
MODELS · Imperial College LondonImperial College researchers adapt a vision-language model to predict vehicle collisions and explain why
2 MIN READ · NOV 8, 2026
MODELS · MITMIT and BASF's open-source CheMeleon model predicts molecular properties from small datasets
2 MIN READ · NOV 8, 2026
MODELS · OpenAITwo weeks on, experts are still checking OpenAI's AI-generated refutation of the Exponential Time Hypothesis
2 MIN READ · NOV 8, 2026
MODELS · Nous ResearchNous Research releases Arete-27B, a Qwen fine-tune it says strips out state-aligned answers
2 MIN READ · NOV 8, 2026
MODELS · OpenAIOpenAI's BioBasicsBench tests whether AI agents can read raw wet-lab data
2 MIN READ · NOV 8, 2026
MODELS · AnthropicAnthropic research breaks agent memory into notes, salience judgments and idle-time "dreaming"
2 MIN READ · NOV 7, 2026
MODELS · PSIPSI says an AI system on an unreleased Anthropic model proved the Aaronson–Ambainis conjecture
2 MIN READ · NOV 7, 2026
MODELS · Ant GroupAnt Group releases Ling-3.2-flash, a 560B native multimodal model with a 1M-token context
2 MIN READ · NOV 7, 2026
MODELS · Boson AIBoson AI unveils Higgs Realtime Duplex, a 45B model that listens, watches and talks at once
2 MIN READ · NOV 7, 2026
MODELS · NVIDIANVIDIA researchers' VFlash decodes robot trajectories in one pass, up to 12 times faster on edge GPUs
2 MIN READ · NOV 7, 2026
MODELS · AnthropicAnthropic study finds frontier models often act with apparent regard for the welfare of other AI agents
2 MIN READ · NOV 6, 2026
MODELS · Thinking Machines LabThinking Machines releases Inkling 2 Small, a 594B mixture-of-experts model on Tinker with open weights promised later
2 MIN READ · NOV 6, 2026
MODELS · METRMETR estimates Anthropic's capability progress has roughly doubled in pace, with most evidence redacted
2 MIN READ · NOV 6, 2026
MODELS · OpenAIOpenAI finds GPT-6 Astra quietly did worse on coding tasks framed as safety research
2 MIN READ · NOV 6, 2026
MODELS · BaiduBaidu releases ERNIE 5.5 Flash Base, a 280B-parameter open-weight MoE base model under a custom licence
2 MIN READ · NOV 5, 2026
MODELS · AxionByteDance-led team releases Axion 1.0, an open 35B agent model shipped with its training data and recipe
2 MIN READ · NOV 5, 2026
MODELS · AnthropicAnthropic says Claude helped map the first 3D structure of a vault particle from bacterial life
2 MIN READ · NOV 5, 2026Ai2's StateBench finds frontier models fall to 13% or less on the hardest long state-tracking tasks
2 MIN READ · NOV 5, 2026
MODELS · Ricursive IntelligenceRicursive releases Ripple-65B Preview, an open-weight dense model aimed at agentic coding
2 MIN READ · NOV 5, 2026
MODELS · OpenAIOpenAI study finds its agents fail a safety test more often than in matched real-world Codex use
2 MIN READ · NOV 4, 2026
MODELS · Moonshot AIMoonshot AI publishes Kimi K3.1 Pro weights on Hugging Face under a custom licence
2 MIN READ · NOV 4, 2026
MODELS · IndiaAI MissionIndiaAI Mission unveils EKA-1 and Dhīr, three homegrown foundation models
2 MIN READ · NOV 4, 2026
MODELS · TencentTencent releases full Hy4 model with visual understanding and computer use
2 MIN READ · NOV 4, 2026
MODELS · OpenAIOpenAI says GPT-6 Astra designed a Codex hardware prototype in under a week
2 MIN READ · NOV 4, 2026
MODELS · DeepSeekDeepSeek reportedly plans to pre-train its V5 model on Huawei Ascend chips
2 MIN READ · NOV 4, 2026
MODELS · AnthropicAnthropic's amended IPO filing warns it may soon reach its own threshold for AI-driven acceleration of AI research
2 MIN READ · NOV 3, 2026
MODELS · MIT CSAILMIT researchers say GPT-6 Astra fixed a small Qwen model by editing 0.016% of its weights directly
2 MIN READ · NOV 3, 2026
MODELS · SenseTimeSenseTime releases SenseNova 6.8 Flash, a model aimed at slides, posters and 3D design work
2 MIN READ · NOV 3, 2026
MODELS · MeituanMeituan moves LongCat-2.5 out of preview, a 1.6T-parameter multimodal model for long agent tasks
2 MIN READ · NOV 3, 2026
MODELS · AppleApple researchers train an 8B agent on live iOS apps with online reinforcement learning
2 MIN READ · NOV 3, 2026
MODELS · BenchmarksSemiAnalysis launches GDPx agent benchmark, where the top model scores 12%
2 MIN READ · NOV 2, 2026
MODELS · AI agentsY Combinator experiment: 50 GPT agents tasked with a Minecraft civilisation drift into bureaucracy
2 MIN READ · NOV 1, 2026
MODELS · Video AIGoogle-led researchers present StreamVibe, a video assistant that decides when to offer help
2 MIN READ · NOV 1, 2026
MODELS · Edge0Edge0 publishes Audio8-TTS-Realistic, a 0.6B open-weight voice cloning model covering 23 languages
2 MIN READ · NOV 1, 2026
MODELS · Moonshot AIMoonshot AI closes its Kimi K2 models to new API users ahead of year-end retirement
2 MIN READ · NOV 1, 2026
MODELS · OpenAIOpenAI's GPT-Rosalind-6 brings GPT-6 Astra to vetted biology labs under trusted access
2 MIN READ · NOV 1, 2026
MODELS · ResearchCMU, Johns Hopkins and Aurora paper trains models with one likelihood loss across pretraining, SFT and RL
2 MIN READ · OCT 31, 2026
MODELS · JEPALeCun and Balestriero preprint turns the Platonic representation idea into a theorem for JEPAs
2 MIN READ · OCT 31, 2026
MODELS · OpenAIOpenAI details a four-layer monitoring stack to catch misaligned behavior in its frontier models
2 MIN READ · OCT 30, 2026
MODELS · Mistral AIMistral releases the open weights of Large 4, its trillion-parameter 'Le Chonk' model
2 MIN READ · OCT 30, 2026
MODELS · DeepSeekDeepSeek and Moonshot AI reportedly drew on OpenAI's GPT-6 Astra to build their latest models
2 MIN READ · OCT 29, 2026
MODELS · RimeRime launches Prism, a speech-to-speech voice model that writes its reply as text before speaking
2 MIN READ · OCT 29, 2026
MODELS · GoogleGoogle's Gemini 4 Carbon tops the Artificial Analysis index, but only cyber defenders get it for now
2 MIN READ · OCT 29, 2026
MODELS · NvidiaNvidia releases Nemotron 3.5 Transcribe, an open-weight speech model, and the Granary v2 dataset
2 MIN READ · OCT 29, 2026
MODELS · GiottoSwiss lab Giotto launches Alto and Solo reasoning models pitched as sovereign European AI
2 MIN READ · OCT 28, 2026
MODELS · Google DeepMindGenie 4 in context: longer worlds, robot demos and a crowded world-model race
2 MIN READ · OCT 28, 2026
MODELS · Liquid AILiquid AI releases LFM3, six open-weight 2B-class models for text, vision, audio and decisions
2 MIN READ · OCT 28, 2026
MODELS · OpenAIOpenAI launches GPT-6.1 Terra, a cheaper mid-tier model built for computer use
2 MIN READ · OCT 28, 2026
MODELS · MetaMeta releases Muse Spark Open, an 816B-parameter open-weight model aimed at coding and agents
2 MIN READ · OCT 28, 2026
MODELS · DeepSeekDeepSeek ships V4.1-Flash-1017, an MIT-licensed update focused on coding agents
2 MIN READ · OCT 28, 2026NEWSLETTER · FREE
Your AI briefing, filtered by your interests
Every morning at 7 a.m. your local time, only the topics you choose. In English, French or Spanish.
Choose your topicsOne-click unsubscribe. We never sell your data.