{"providers":[{"id":"anthropic","name":"Anthropic","country":"US","website":"https://www.anthropic.com","description":"AI safety company behind the Claude model family."},{"id":"openai","name":"OpenAI","country":"US","website":"https://openai.com","description":"Creator of the GPT series, ChatGPT and the open-weight gpt-oss models."},{"id":"google","name":"Google","country":"US","website":"https://deepmind.google","description":"Google DeepMind builds the Gemini, Gemma and Nano Banana model families."},{"id":"meta","name":"Meta","country":"US","website":"https://ai.meta.com","description":"Publisher of the open-weight Llama models, the Muse model family and the V-JEPA world models from FAIR."},{"id":"xai","name":"xAI (SpaceXAI)","country":"US","website":"https://x.ai","description":"Builder of the Grok models, now operating as SpaceXAI."},{"id":"alibaba","name":"Alibaba (Qwen)","country":"CN","website":"https://qwen.ai","description":"The Qwen team ships both proprietary Max models and widely used open-weight models."},{"id":"deepseek","name":"DeepSeek","country":"CN","website":"https://www.deepseek.com","description":"Research lab known for efficient, open-weight mixture-of-experts models."},{"id":"moonshot","name":"Moonshot AI","country":"CN","website":"https://www.moonshot.ai","description":"Creator of the Kimi assistant and open-weight Kimi K-series models."},{"id":"zai","name":"Z.ai","country":"CN","website":"https://z.ai","description":"Formerly Zhipu AI; publishes the open-weight GLM family."},{"id":"xiaomi","name":"Xiaomi","country":"CN","website":"https://mimo.xiaomi.com","description":"Consumer-electronics company publishing the open MiMo model family."},{"id":"mistral","name":"Mistral AI","country":"FR","website":"https://mistral.ai","description":"European lab offering open-weight and commercial models."},{"id":"minimax","name":"MiniMax","country":"CN","website":"https://www.minimax.io","description":"Multimodal AI company with open-weight long-context models."},{"id":"stepfun","name":"StepFun","country":"CN","website":"https://www.stepfun.com","description":"Lab building the Step family of multimodal agentic models."},{"id":"cohere","name":"Cohere","country":"CA","website":"https://cohere.com","description":"Enterprise-focused models for agents, RAG and search."},{"id":"nvidia","name":"NVIDIA","country":"US","website":"https://www.nvidia.com/en-us/ai/","description":"Publishes the open Nemotron language models, Cosmos world foundation models and Isaac GR00T robotics models."},{"id":"inception","name":"Inception","country":"US","website":"https://www.inceptionlabs.ai","description":"Pioneer of diffusion LLMs (dLLMs) for very high throughput."},{"id":"reflection","name":"Reflection AI","country":"US","website":"https://reflection.ai","description":"Builds open-weight frontier models for coding and agents."},{"id":"ibm","name":"IBM","country":"US","website":"https://www.ibm.com/granite","description":"Publishes the Apache-licensed Granite enterprise models."},{"id":"amazon","name":"Amazon","country":"US","website":"https://aws.amazon.com/ai/generative-ai/nova/","description":"Amazon Nova models, available through AWS Bedrock."},{"id":"typesafe","name":"TypeSafe AI","country":"US","website":"https://typesafe.ai","description":"San Francisco lab building System One models — fast, calibrated decision models for software, starting with Jev."},{"id":"bytedance","name":"ByteDance Seed","country":"CN","website":"https://seed.bytedance.com","description":"ByteDance’s foundation-model team: Seed LLMs, Seedream images and Seedance video."},{"id":"meituan","name":"Meituan (LongCat)","country":"CN","website":"https://longcat.ai","description":"Publishes the open-weight LongCat mixture-of-experts models."},{"id":"inclusionai","name":"inclusionAI (Ant Group)","country":"CN","website":"https://github.com/inclusionAI","description":"Ant Group’s open AI lab behind the Ling and Ring model families."},{"id":"liquid","name":"Liquid AI","country":"US","website":"https://www.liquid.ai","description":"Builds efficient Liquid Foundation Models (LFMs) for on-device and edge use."},{"id":"microsoft","name":"Microsoft AI","country":"US","website":"https://microsoft.ai","description":"Microsoft’s in-house MAI models for images, voice and transcription."},{"id":"black-forest-labs","name":"Black Forest Labs","country":"DE","website":"https://bfl.ai","description":"Creators of the FLUX image and video generation models."},{"id":"kuaishou","name":"Kuaishou (Kling)","country":"CN","website":"https://klingai.com","description":"Maker of the Kling video generation models."},{"id":"lightricks","name":"Lightricks","country":"IL","website":"https://www.lightricks.com","description":"Publishes the open-weight LTX video generation models."},{"id":"elevenlabs","name":"ElevenLabs","country":"US","website":"https://elevenlabs.io","description":"Voice AI company: text-to-speech, Scribe transcription, music and voice agents."},{"id":"cartesia","name":"Cartesia","country":"US","website":"https://cartesia.ai","description":"Real-time voice models (Sonic) built on state-space model research."},{"id":"inworld","name":"Inworld","country":"US","website":"https://inworld.ai","description":"Real-time voice and character AI for interactive applications."},{"id":"assemblyai","name":"AssemblyAI","country":"US","website":"https://www.assemblyai.com","description":"Speech-to-text and speech understanding APIs."},{"id":"voyage","name":"Voyage AI (MongoDB)","country":"US","website":"https://www.voyageai.com","description":"Embedding and reranking models for retrieval; part of MongoDB."},{"id":"perplexity","name":"Perplexity","country":"US","website":"https://www.perplexity.ai","description":"AI search company; also publishes open embedding and decision models."},{"id":"physical-intelligence","name":"Physical Intelligence","country":"US","website":"https://www.physicalintelligence.company","description":"Robot foundation-model lab behind the open π0 / π0.5 VLA models."},{"id":"tencent","name":"Tencent","country":"CN","website":"https://hunyuan.tencent.com","description":"Builder of the Hy (formerly Hunyuan) language and image model families."},{"id":"upstage","name":"Upstage","country":"KR","website":"https://www.upstage.ai","description":"Korean lab behind the Solar language models and Solar Decide."},{"id":"jetbrains","name":"JetBrains","country":"CZ","website":"https://www.jetbrains.com","description":"IDE maker that trains the open Mellum models for coding agents."},{"id":"cloudflare","name":"Cloudflare","country":"US","website":"https://www.cloudflare.com","description":"Edge network whose Workers AI team builds the open-source Clef decision models."},{"id":"kandinsky-lab","name":"Kandinsky Lab (Sber)","country":"RU","website":"https://github.com/kandinskylab","description":"Sber’s generative lab behind the open-weight Kandinsky image and video models."},{"id":"noiz","name":"Noiz AI","country":"HK","website":"https://noiz.ai","description":"Audio-visual AI lab that built the Helix World model with HKUST."},{"id":"thinking-machines","name":"Thinking Machines Lab","country":"US","website":"https://thinkingmachines.ai","description":"AI lab that publishes the open-weight Inkling models and the Tinker fine-tuning API."},{"id":"krea","name":"Krea","country":"US","website":"https://www.krea.ai","description":"Creative AI company that trains the Krea image models."},{"id":"recraft","name":"Recraft","country":"US","website":"https://www.recraft.ai","description":"Design-focused image generation lab known for vector and brand-consistent output."},{"id":"runway","name":"Runway","country":"US","website":"https://runwayml.com","description":"Creative AI company behind the Gen and Aleph video models."}],"counts":{"models":149,"tools":264,"providers":45},"dataUpdatedAt":"2026-10-30","sources":[{"name":"LLM Stats","url":"https://llm-stats.com/leaderboards/llm-leaderboard","usedFor":"Reasoning, coding, agentic, math, tool-calling and vision composites; SWE-bench Pro (vendor-reported); knowledge cutoffs and parameter counts; dated on each page"},{"name":"OpenRouter Data API (CC BY 4.0)","url":"https://openrouter.ai/rankings","usedFor":"Usage share and rank (30 days), most-used tasks, app usage ranks for tools, and OpenRouter’s own GPQA Diamond and τ-bench evals; dated on each page"},{"name":"Arena leaderboard dataset (CC BY 4.0)","url":"https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset","usedFor":"Human-preference ratings: text, WebDev, text-to-image, image editing, text-to-video and image-to-video; dated on each page"},{"name":"Epoch AI Benchmarking Hub (CC BY 4.0)","url":"https://epoch.ai/benchmarks","usedFor":"Epoch Capabilities Index (ECI) for language models; dated on each page"},{"name":"Model makers & Hugging Face model cards","url":"https://huggingface.co","usedFor":"Prices, context windows, output limits, release dates, modalities, licenses and architectures, cited per model; descriptions are written by SI Index"}]}