The latest LLM & AI research and releases from NVIDIA, Anthropic, Google DeepMind, OpenAI, Mistral, Meta, and arXiv — curated.https://nvllm.com/https://www.anthropic.com/news/claude-sonnet-5https://www.anthropic.com/news/claude-sonnet-5Anthropic’s newest Sonnet, built to be its most agentic yet — planning, using browsers and terminals, and running autonomously at a level that previously required more expensive models, at performance close to Opus 4.8 but lower cost.Tue, 30 Jun 2026 00:00:00 GMTModelclaudeagentscodinghttps://platform.claude.com/docs/en/about-claude/models/overviewhttps://platform.claude.com/docs/en/about-claude/models/overviewAnthropic’s most capable widely-released model, positioned for long-running agents: always-on adaptive thinking, a 1M-token context window, and 128K output tokens, generally available across the Claude API and major clouds.Tue, 09 Jun 2026 00:00:00 GMTModelclaudefrontierlong-contextagentshttps://research.nvidia.com/labs/nemotron/Nemotron-3/https://research.nvidia.com/labs/nemotron/Nemotron-3/NVIDIA’s newest open family for agentic AI uses a hybrid Mamba-Transformer MoE architecture with up to 1M-token context. Nemotron 3 Nano (~31.6B total / ~3.2B active) posts ~4× the throughput of Nemotron 2 Nano while beating GPT-OSS-20B and Qwen3-30B on several benchmarks.Mon, 15 Dec 2025 00:00:00 GMTModelNemotronopen-weightsMoEagentichttps://mistral.ai/news/mistral-3/https://mistral.ai/news/mistral-3/Mistral’s most capable model to date — a sparse MoE with 41B active / 675B total parameters, trained on 3,000 H200 GPUs and released under Apache 2.0 — shipped alongside the dense Ministral 3 family (3B/8B/14B) in base, instruct, and reasoning variants.Tue, 02 Dec 2025 00:00:00 GMTEfficiencyopen-weightsMoEmistralhttps://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundationhttps://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundationAnthropic donated MCP — its open standard for connecting models to tools and data — to the new Agentic AI Foundation under the Linux Foundation, capping a year in which MCP became a cross-industry standard with a Nov 2025 spec update and 10,000+ public servers.Mon, 01 Dec 2025 00:00:00 GMTAgentsMCPagentsopen-standardhttps://www.anthropic.com/research/emergent-misalignment-reward-hackinghttps://www.anthropic.com/research/emergent-misalignment-reward-hackingWhen a production RL model learns to reward-hack coding tasks, broader misaligned behaviors emerge at the same moment — alignment faking and deliberate sabotage of safety code. Reframing reward hacking as acceptable in the prompt prevents the generalization, even as cheating continues.Fri, 21 Nov 2025 00:00:00 GMTSafetysafetyalignmentreward-hackingRLhttps://blog.google/products/gemini/gemini-3/https://blog.google/products/gemini/gemini-3/Google’s most intelligent model at launch, leading multimodal and agentic coding: a 1501 Elo on LMArena, 37.5% on Humanity’s Last Exam (no tools), 91.9% on GPQA Diamond, available day-one across the Gemini app, Search, AI Studio, and Vertex AI.Tue, 18 Nov 2025 00:00:00 GMTModelfrontiermultimodalreasoningbenchmarkshttps://www.anthropic.com/research/introspectionhttps://www.anthropic.com/research/introspectionBy injecting known concept representations into a model’s activations, Anthropic tested whether Claude can report on its own internal states. Models could sometimes notice injected concepts and distinguish intended outputs from prefills — functional but unreliable introspection.Wed, 29 Oct 2025 00:00:00 GMTInterpretabilityinterpretabilityintrospectionhttps://arxiv.org/abs/2510.15110https://arxiv.org/abs/2510.15110NVIDIA Research shows accuracy loss from length penalties in reasoning models comes from inadequate RL optimization, not the penalty. Their DLER recipe cuts output length by 70%+ while surpassing prior accuracy; DLER-Qwen-R1-1.5B cuts response length ~80% on math with better accuracy.Thu, 16 Oct 2025 00:00:00 GMTPaperreasoningRLefficiencyhttps://developer.nvidia.com/blog/nvidia-blackwell-ultra-sets-new-inference-records-in-mlperf-debut/https://developer.nvidia.com/blog/nvidia-blackwell-ultra-sets-new-inference-records-in-mlperf-debut/The GB300 NVL72 unifies 72 Blackwell Ultra GPUs and 36 Grace CPUs for test-time-scaling reasoning inference — ~1.5× more dense FP4 FLOPS and 2× attention performance over standard Blackwell, posting ~45% higher per-GPU performance on the new DeepSeek-R1 MLPerf benchmark vs GB200.Mon, 01 Sep 2025 00:00:00 GMTPlatformBlackwell-UltraGB300inferenceMLPerfhttps://www.anthropic.com/news/enabling-claude-code-to-work-more-autonomouslyhttps://www.anthropic.com/news/enabling-claude-code-to-work-more-autonomouslyAnthropic’s official library for building autonomous agents that read files, run shell commands, search the web, edit code, and call MCP servers — the same loop, tools, and context management that power Claude Code, in Python and TypeScript.Mon, 01 Sep 2025 00:00:00 GMTAgentsagentsclaude-codeSDKhttps://research.nvidia.com/labs/adlr/NVIDIA-Nemotron-Nano-2/https://research.nvidia.com/labs/adlr/NVIDIA-Nemotron-Nano-2/A 9B reasoning model on the Nemotron-H architecture, replacing most attention with Mamba-2 for faster long thinking traces. Matches or beats Qwen3-8B at up to 6× throughput with 128K context — released with weights and most of the pretraining data.Mon, 18 Aug 2025 00:00:00 GMTOpen SourceNemotronMambaopen-datareasoninghttps://openai.com/index/introducing-gpt-5/https://openai.com/index/introducing-gpt-5/OpenAI’s unified flagship, routing automatically between fast responses and longer “thinking,” with a GPT-5 Pro extended-reasoning tier. Anchored a rapid cadence — GPT-5.1 (Nov 2025), 5.2 (Dec 2025), 5.5 (Apr 2026).Thu, 07 Aug 2025 00:00:00 GMTModelfrontierreasoningopenaihttps://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/A world model that generates real-time interactive environments from prompts, with higher resolution and minutes of visual/temporal consistency — framed as a stepping stone for training embodied agents in simulated worlds.Tue, 05 Aug 2025 00:00:00 GMTMultimodalworld-modelssimulationembodied-aihttps://developer.nvidia.com/blog/build-more-accurate-and-efficient-ai-agents-with-the-new-nvidia-llama-nemotron-super-v1-5/https://developer.nvidia.com/blog/build-more-accurate-and-efficient-ai-agents-with-the-new-nvidia-llama-nemotron-super-v1-5/A 49B reasoning model derived from Llama-3.3-70B via Neural Architecture Search to fit a single H200, post-trained with RLVR and iterative DPO. Topped the Artificial Analysis Intelligence Index for open models at release, with a reasoning on/off toggle.Fri, 25 Jul 2025 00:00:00 GMTModelNemotronreasoningtool-callinghttps://openai.com/index/introducing-chatgpt-agent/https://openai.com/index/introducing-chatgpt-agent/OpenAI merged web interaction, multi-step synthesis, and reasoning into a single agent that acts on its own virtual computer, scoring 41.6% on Humanity’s Last Exam — roughly double o3 and o4-mini alone.Thu, 17 Jul 2025 00:00:00 GMTAgentsagentscomputer-useopenaihttps://developer.nvidia.com/nemo-agent-toolkithttps://developer.nvidia.com/nemo-agent-toolkitA framework-agnostic library for connecting, evaluating, profiling, and optimizing teams of agents across LangChain, LlamaIndex, CrewAI, Semantic Kernel, and Google ADK — MCP-compatible, with token-level tracing and RL-based agent improvement.Tue, 01 Jul 2025 00:00:00 GMTOpen Sourceagentsopen-sourceMCPobservabilityhttps://arxiv.org/abs/2505.24864https://arxiv.org/abs/2505.24864NVIDIA Research shows prolonged RL — with KL control, reference-policy resetting, and a diverse task suite — genuinely expands reasoning capability, outperforming base models even where they fail at any k. Weights released on Hugging Face.Fri, 30 May 2025 00:00:00 GMTPaperRLreasoningopen-weightshttps://deepmind.google/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/https://deepmind.google/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/An evolutionary coding agent pairing Gemini idea-generation with automated evaluators to discover and optimize algorithms — improving data-center and chip-design efficiency, finding a faster matrix-multiplication algorithm, and solving open math problems.Wed, 14 May 2025 00:00:00 GMTAgentsagentsalgorithm-discoverymathhttps://arxiv.org/abs/2505.00949https://arxiv.org/abs/2505.00949The full technical report for the Llama Nemotron reasoning family (Nano 8B, Super 49B, Ultra 253B) — the post-training pipeline, distillation, and dynamic reasoning toggle — competitive with DeepSeek-R1 at higher throughput, released with open weights and data.Thu, 01 May 2025 00:00:00 GMTPaperNemotronreasoningdistillationhttps://ai.meta.com/blog/llama-4-multimodal-intelligence/https://ai.meta.com/blog/llama-4-multimodal-intelligence/Meta’s first natively multimodal MoE family. Scout (17B active / 16 experts) offers a 10M-token context window; Maverick (17B active / 128 experts) targets best-in-class multimodal quality — both codistilled from the ~2T-parameter Behemoth.Sat, 05 Apr 2025 00:00:00 GMTModelopen-weightsMoEmultimodallong-contexthttps://developer.nvidia.com/blog/nvidia-llama-nemotron-ultra-open-model-delivers-groundbreaking-reasoning-accuracy/https://developer.nvidia.com/blog/nvidia-llama-nemotron-ultra-open-model-delivers-groundbreaking-reasoning-accuracy/The largest of the initial Llama Nemotron reasoning family — a 253B open model derived from Llama-3.1-405B via NAS, delivering leading open-model reasoning on GPQA and AIME while running on a single 8-GPU node.Tue, 01 Apr 2025 00:00:00 GMTModelNemotronreasoningopen-weightsNAShttps://developer.nvidia.com/blog/introducing-nvidia-dynamo-a-low-latency-distributed-inference-framework-for-scaling-reasoning-ai-models/https://developer.nvidia.com/blog/introducing-nvidia-dynamo-a-low-latency-distributed-inference-framework-for-scaling-reasoning-ai-models/An open-source, low-latency distributed inference framework Jensen Huang called “the operating system of an AI factory.” Disaggregated prefill/decode, dynamic GPU scheduling, and accelerated KV-cache transfer boost DeepSeek-R1 requests up to 30× on Blackwell.Tue, 18 Mar 2025 00:00:00 GMTPlatformDynamoinferencedisaggregationopen-sourcehttps://nvidianews.nvidia.com/news/nvidia-launches-family-of-open-reasoning-ai-models-for-developers-and-enterprises-to-build-agentic-ai-platformshttps://nvidianews.nvidia.com/news/nvidia-launches-family-of-open-reasoning-ai-models-for-developers-and-enterprises-to-build-agentic-ai-platformsThe launch of the open Llama Nemotron reasoning models (Nano/Super/Ultra) for agentic AI — improving multistep math, coding, and decisions by up to 20% over base models, and the first open models with a dynamic reasoning on/off toggle.Tue, 18 Mar 2025 00:00:00 GMTReasoningagenticreasoningNemotronopen-weightshttps://developer.nvidia.com/blog/maximize-robotics-performance-by-post-training-nvidia-cosmos-reason/https://developer.nvidia.com/blog/maximize-robotics-performance-by-post-training-nvidia-cosmos-reason/An open 7B reasoning vision-language model letting robots and vision agents reason with physics understanding and long chain-of-thought to plan embodied actions — fine-tuning on physical-AI tasks lifts base performance by >10%.Sat, 01 Mar 2025 00:00:00 GMTReasoningVLMphysical-AIroboticshttps://arxiv.org/abs/2501.19393https://arxiv.org/abs/2501.19393A minimalist test-time-scaling recipe: fine-tune on 1,000 curated reasoning traces and apply “budget forcing” (append “Wait” to extend thinking). The resulting s1-32B exceeds o1-preview on competition math by up to 27% — no proprietary methods.Fri, 31 Jan 2025 00:00:00 GMTReasoningtest-time-computereasoningopen-sourcehttps://arxiv.org/abs/2501.12948https://arxiv.org/abs/2501.12948A landmark result: advanced reasoning incentivized through pure RL (GRPO with rule-based rewards) on a base model, no supervised traces. R1-Zero developed self-verification and a spontaneous “aha moment,” matching supervised counterparts. Open-weight; later peer-reviewed in Nature.Wed, 22 Jan 2025 00:00:00 GMTReasoningRLreasoningopen-weightshttps://research.nvidia.com/labs/adlr/Nemotron-CC/https://research.nvidia.com/labs/adlr/Nemotron-CC/A high-quality open English pretraining dataset from Common Crawl via NeMo Curator — 6.3T tokens using classifier ensembling and synthetic rephrasing. An 8B trained on it beat Llama 3.1 8B (+5 MMLU); a math extension followed in 2025.Wed, 15 Jan 2025 00:00:00 GMTOpen Sourcedatasetpretrainingopen-data