Frontier research laboratories driving machine intelligence forward. Each organization maintains a primary foundation model alongside focused experimental checkpoints.
Anthropic
(San Francisco, CA)Creator of the Claude family; pioneers of adaptive thinking, Enterprise Frontier Safeguards (EFS), and agentic knowledge work.
Released Sept 1, 2026; Anthropic's reigning flagship for agentic coding and knowledge work. Features a 75% reduction in cache read costs ($0.25/M tokens) and 52.6% on Terminal-Bench-Science.
Anthropic's reigning foundation flagship is also its newest deployed model.
OpenAI
(San Francisco, CA)Pioneered unified foundation models with GPT-5.6 (Sol, Terra, Luna) and the Astra critical-capability cybersecurity model.
OpenAI's primary general-purpose flagship model. Combines native fast execution, multi-agent orchestration, and 1.05M context for production software development.
Announced Sept 1, 2026; the first LLM to meet OpenAI's 'Critical' cybersecurity threshold under its Preparedness Framework, capable of identifying zero-day exploits.
Google DeepMind
(London, UK / Mountain View, CA)Creators of the Gemini 3 family, leading ultra-fast agentic coding, multi-step workflows, and 1M+ token real-time multimodal streaming.
Released Sept 2, 2026; Google DeepMind's premier frontier workhorse powering advanced agentic coding loops, recursive self-correction, and 1M token real-time multimodal streaming.
Google DeepMind's reigning foundation flagship is also its newest deployed model.
DeepSeek
(Hangzhou, China)Open-source frontier champion famous for DeepSeek-V4 massive MoE architecture (1.6T parameters) and DSpark speculative decoding.
DeepSeek's primary 1.6T parameter powerhouse with 49B activated per token, configurable thinking budget, and 1M context.
Open-sourced under MIT license on August 30, 2026. 305B parameter multimodal vision-language model with native document understanding.
xAI
(San Francisco, CA / Memphis, TN)Elon Musk's lab powered by the Memphis Colossus supercluster, creators of Grok 4.6 with 1.5T parameters and upcoming Grok 4.7.
Released August 12, 2026; xAI's smartest model with frontier performance in coding and autonomous agents, integrated natively into Cursor and Grok Build.
xAI's reigning foundation flagship is also its newest deployed model.
Alibaba Cloud (Qwen)
(Hangzhou, China)Massive open and proprietary foundation ecosystem, leading global charts with the Qwen3.8 series and 2.4T open MoE models.
The largest open-weight MoE model in existence. 2.4 Trillion parameters with 95B activated per token and 1M context, available on HuggingFace and ModelScope.
Released August 26, 2026; combines visual document understanding, fast agentic workflows, and 1M context with ultra-cheap $0.15/$0.47 pricing.
Meta AI
(Menlo Park, CA)Global open-source foundation leader with Llama 4 (Maverick, Scout) and real-time speech foundation model Muse Voice Transcribe.
Meta's flagship open-weights foundation model. 128-expert MoE architecture with 1M token context, powering enterprise on-premise deployments and community fine-tuning.
Announced Sept 1, 2026; streaming speech-to-text foundation model supporting 70+ languages, 20+ voice diarization, and multilingual code-switching.
Mistral AI
(Paris, France)European frontier lab building high-efficiency dense and MoE models including Mistral Medium 3.5, Mistral Small 4, and Devstral 2.
Dense 128B multimodal instruction-following model with native text and image understanding, tuned for European enterprise compliance.
Mistral AI's reigning foundation flagship is also its newest deployed model.
Cohere
(Toronto, Canada)Enterprise RAG and agentic coding specialists behind North Mini Code (30B MoE) and open-weight Command A (111B).
Cohere's primary flagship model optimized for business agentic search, long-document question answering, and multi-step tool use.
30B MoE agentic coding model specifically tuned for multi-file workspace inspection and test suites, offered free for developers.