 01.AIStartup lab · Beijing, ChinaBeijing, ChinaKai-Fu Lee's lab behind the open Yi models, now focused on enterprise AI products. | Kai-Fu Lee's lab behind the open Yi models, now focused on enterprise AI products. | 2023 | Yi-34B | 0 |
 AI21 LabsStartup lab · Tel Aviv, IsraelTel Aviv, IsraelJamba, the first production-scale hybrid of Transformer and Mamba layers. | Jamba, the first production-scale hybrid of Transformer and Mamba layers. | 2017 | None | 0 |
 AnthropicStartup lab · San Francisco, United StatesSan Francisco, United StatesThe Claude models, Constitutional AI and interpretability research on what happens inside models. | The Claude models, Constitutional AI and interpretability research on what happens inside models. | 2021 | Claude Fable 5.1 | 4 |
 Black Forest LabsStartup lab · Freiburg, GermanyFreiburg, GermanyThe FLUX image models, from the researchers behind latent diffusion and Stable Diffusion. | The FLUX image models, from the researchers behind latent diffusion and Stable Diffusion. | 2024 | FLUX | 0 |
 Cohere and Cohere LabsStartup lab · Toronto, CanadaToronto, CanadaEnterprise models (Command, Embed, Rerank) and Cohere Labs' open multilingual Aya research. | Enterprise models (Command, Embed, Rerank) and Cohere Labs' open multilingual Aya research. | 2019 | Command A | 0 |
 DeepSeekStartup lab · Hangzhou, ChinaHangzhou, ChinaFrontier-class open-weight models trained cheaply: DeepSeek-V3's efficient MoE and R1's open reasoning recipe. | Frontier-class open-weight models trained cheaply: DeepSeek-V3's efficient MoE and R1's open reasoning recipe. | 2023 | DeepSeek V4.1 Flash | 2 |
 DuckLabs (DuckDB Labs)Startup lab · NetherlandsNetherlandsThe company of DuckDB's creators, the in-process analytics database born at CWI; Amazon agreed to buy it in 2026. | The company of DuckDB's creators, the in-process analytics database born at CWI; Amazon agreed to buy it in 2026. | n/a | None | 0 |
 Hugging FaceStartup lab · New York, United StatesNew York, United StatesThe Hub for open models and datasets and the Transformers library; also BLOOM, StarCoder, FineWeb and SmolLM. | The Hub for open models and datasets and the Transformers library; also BLOOM, StarCoder, FineWeb and SmolLM. | 2016 | None | 0 |
 Liquid AIStartup lab · Cambridge, Massachusetts, United StatesCambridge, Massachusetts, United StatesLiquid foundation models (LFM) for devices, built on liquid neural network research from MIT CSAIL. | Liquid foundation models (LFM) for devices, built on liquid neural network research from MIT CSAIL. | 2023 | None | 0 |
 MiniMaxStartup lab · Shanghai, ChinaShanghai, ChinaLightning attention for million-token context in open models, plus Hailuo video and speech models. | Lightning attention for million-token context in open models, plus Hailuo video and speech models. | 2021 | None | 0 |
 Mistral AIStartup lab · Paris, FranceParis, FranceEurope's leading model maker: efficient open-weight models, Mixtral's mixture of experts, and Le Chat. | Europe's leading model maker: efficient open-weight models, Mixtral's mixture of experts, and Le Chat. | 2023 | Mistral Medium 3 | 0 |
 Moonshot AIStartup lab · Beijing, ChinaBeijing, ChinaThe Kimi models: long context early, then the open-weight Kimi K2 trained with the Muon optimizer. | The Kimi models: long context early, then the open-weight Kimi K2 trained with the Muon optimizer. | 2023 | Kimi K3 | 0 |
 Nous ResearchStartup lab · United StatesUnited StatesThe open Hermes fine-tunes, YaRN context extension, and decentralized training over the internet. | The open Hermes fine-tunes, YaRN context extension, and decentralized training over the internet. | n/a | None | 0 |
 OpenAIStartup lab · San Francisco, United StatesSan Francisco, United StatesThe GPT models and ChatGPT, and the scaling results that set off the large language model race. | The GPT models and ChatGPT, and the scaling results that set off the large language model race. | 2015 | GPT-6 Astra | 5 |
 Physical IntelligenceStartup lab · San Francisco, United StatesSan Francisco, United StatesGeneral-purpose robot foundation models: the pi-zero vision-language-action family, open-sourced as openpi. | General-purpose robot foundation models: the pi-zero vision-language-action family, open-sourced as openpi. | 2024 | None | 0 |
 RekaStartup lab · Sunnyvale, United StatesSunnyvale, United StatesNatively multimodal models (Core, Flash, Edge) from former DeepMind and Meta researchers. | Natively multimodal models (Core, Flash, Edge) from former DeepMind and Meta researchers. | 2022 | None | 0 |
 Safe Superintelligence (SSI)Startup lab · Palo Alto, United StatesPalo Alto, United StatesIlya Sutskever's lab with one goal and one product, a safe superintelligence; it has published nothing yet. | Ilya Sutskever's lab with one goal and one product, a safe superintelligence; it has published nothing yet. | 2024 | None | 0 |
 Sakana AIStartup lab · Tokyo, JapanTokyo, JapanNature-inspired methods: evolutionary model merging and The AI Scientist, from a Transformer co-author. | Nature-inspired methods: evolutionary model merging and The AI Scientist, from a Transformer co-author. | 2023 | None | 0 |
 Stability AIStartup lab · London, United KingdomLondon, United KingdomStable Diffusion, the open image model release that opened up generative imaging in 2022. | Stable Diffusion, the open image model release that opened up generative imaging in 2022. | 2019 | None | 0 |
 Thinking Machines LabStartup lab · San Francisco, United StatesSan Francisco, United StatesMira Murati's lab: Tinker, a fine-tuning API, and open research notes on reproducible inference. | Mira Murati's lab: Tinker, a fine-tuning API, and open research notes on reproducible inference. | 2025 | Inkling | 0 |
 World LabsStartup lab · San Carlos, United StatesSan Carlos, United StatesFei-Fei Li's spatial intelligence company: models that generate explorable 3D worlds (Marble). | Fei-Fei Li's spatial intelligence company: models that generate explorable 3D worlds (Marble). | n/a | None | 0 |
 Z.ai (Zhipu AI)Startup lab · Beijing, ChinaBeijing, ChinaThe GLM models, spun out of Tsinghua University's KEG lab; GLM-4.5 and later are open weights. | The GLM models, spun out of Tsinghua University's KEG lab; GLM-4.5 and later are open weights. | 2019 | GLM-5.3 | 0 |
