Renaissance Labs

Open Labs Atlas

The open frontier of intelligence.

A world map of independent labs releasing open weights, open code and open data, with a page for each.

227labs
54countries
7regions
⤢
227 labs
Academic lab
Espoo, Finland

Aalto Speech Recognition

Aalto University's speech recognition group, whose open Morfessor and AaltoASR tools have long supported speech and language technology for morphologically rich languages.

speechmorphology
Academic lab
Darmstadt, Germany

AdapterHub

AdapterHub maintains the open Adapters library, a modular, parameter-efficient fine-tuning toolkit built on Hugging Face Transformers.

parameter-efficient fine-tuningmodular learningNLP
Academic lab
Berkeley, United States

Agentica

A Berkeley Sky Computing Lab initiative that publishes open reinforcement learning recipes, models and the rLLM framework for training language agents.

reinforcement learningreasoningcoding
Nonprofit
Singapore, Singapore

AI Singapore

Singapore's national AI programme, which builds applied AI capability and the open SEA-LION language models for Southeast Asia.

language modelsSoutheast Asian languages
Institute
Gothenburg, Sweden

AI Sweden

Sweden's national centre for applied AI, a public and private partnership that also published the GPT-SW3 Nordic language models.

language modelsSwedishNordic languages
Independent company
Tel Aviv, Israel

AI21 Labs

An Israeli enterprise AI company that publishes the hybrid SSM-Transformer Jamba models as open weights alongside commercial products.

language modelsenterprise
Nonprofit
Chennai, India

AI4Bharat

An IIT Madras research lab building open datasets, models and tools for Indian languages in translation, speech and language modelling.

datasetsspeechtranslation
Academic lab
Riga, Latvia

AiLab

The AI Laboratory at the University of Latvia builds open Latvian language resources, from the Tezaurs lexicon to a multilayer corpus and Latvian BERT models.

Latvian NLPlexiconslanguage models
Academic lab
Darmstadt, Germany

AIML Lab

The AI and Machine Learning Lab at TU Darmstadt releases open safety tooling and reasoning research, including LlavaGuard and Safe Latent Diffusion.

AI safetyvision-language
Independent company
Heidelberg, Germany

Aleph Alpha

Heidelberg language-model company that built sovereign European models and released part of its research as open weights, now being merged into Cohere.

language modelsGermantokenization
Nonprofit
Aarhus, Denmark

Alexandra Institute

Danish non-profit research and technology institute whose open resources include DaNLP and the ScandEval benchmark, now EuroEval.

Danish NLPevaluationresponsible AI
Nonprofit
Seattle, United States

Allen Institute for AI

A Seattle nonprofit research institute founded by Paul Allen that builds fully open language models, datasets and training recipes.

language modelsdatasetspost-training
Institute
Paris, France

ALMAnaCH

Inria Paris research team behind the CamemBERT family of open French language models and work in computational humanities.

NLPFrenchlanguage models
Independent company
San Francisco, United States

Answer.AI

A public benefit AI R&D lab that turns research advances into open-source libraries, models and tools for developers.

encoder modelsdeveloper toolsretrieval
Institute
London, United Kingdom

Apollo Research

London-based AI safety organisation that evaluates frontier models for scheming and publishes open evaluation and interpretability work.

AI safetyevaluationsinterpretability
Nonprofit
Palo Alto, United States

Arc Institute

A nonprofit biomedical research institute that builds open AI models for biology, including Evo 2 and virtual cell models.

biologyfoundation modelsdatasets
Nonprofit
San Francisco, United States

ARC Prize Foundation

A nonprofit that runs the ARC-AGI benchmarks and prize competitions to measure and encourage general reasoning in AI systems.

evaluationreasoning
Independent company
San Francisco, United States

Arcee AI

A San Francisco model lab that trains open-weight models and maintains MergeKit, a widely used toolkit for merging open language models.

language modelsmodel mergingdistillation
Independent company
Gilching, Germany

audEERING

Munich-area audio AI company and maker of the openSMILE toolkit and open speech emotion and age-gender models.

speech emotion recognitionaudio analysis
Academic lab
Freiburg, Germany

AutoML Groups

The AutoML groups of Freiburg and Hannover, creators of auto-sklearn, SMAC3 and Auto-PyTorch.

AutoML
Nonprofit
Beijing, China

BAAI

A Beijing nonprofit AI research academy behind the Wudao models, the FlagOpen ecosystem and the BGE embedding models.

datasetslanguage modelsembodied AI
Independent company
Mountain View, United States

Bespoke Labs

A data research lab that builds the Curator data tooling, RL environments and benchmarks, and co-leads the OpenThoughts open reasoning data project.

synthetic datapost-trainingfact-checking
Academic lab
Tubingen, Germany

Bethge Lab

Tübingen lab studying robustness and human-like vision, known for the Foolbox toolbox and open robustness benchmarks.

robustnessvision
Institute
Mumbai, India

BharatGen

An India-wide consortium anchored at IIT Bombay that releases open-weight text, speech and vision models for Indian languages.

language modelsevaluationIndic languages
Community collective
New York, United States

BigCode

Open scientific collaboration run by Hugging Face and ServiceNow Research that released The Stack and the StarCoder family of code models.

language modelscodedatasets
Community collective
Paris, France

BigScience

A year-long open research workshop coordinated by Hugging Face that produced the 176B-parameter BLOOM multilingual model.

language modelsdatasetsdistributed
Independent company
Paris, France

Bioptimus

A Paris startup building foundation models for pathology and biology, including the openly released H-Optimus pathology models.

sciencebiologypathology
Independent company
Freiburg, Germany

Black Forest Labs

A Freiburg lab founded by the creators of Stable Diffusion that builds the FLUX image models, with open-weight and API tiers.

image generationvisiondiffusion
Independent company
Cambridge, United States

Boltz

A public benefit company growing out of MIT research that develops biomolecular models, with open Boltz-2 and BoltzGen alongside proprietary ones.

biomolecular modellingstructure predictiondrug discovery
Independent company
San Francisco, United States

Browser Use

An open-source library and company for AI browser agents, with hosted agents and stealth browsers.

agentsbrowser automation
Institute
Barcelona, Spain

BSC Language Technologies Lab

The Barcelona Supercomputing Center language technologies group that trains the open Salamandra and ALIA models for Iberian and European languages.

language modelsSpanishCatalan
Academic lab
Dhaka, Bangladesh

BUET CSE NLP Group

A BUET research group on low-resource Bangla NLP that releases BanglaBERT, BanglaT5 and the XL-Sum summarisation dataset.

BanglaNLPdatasets
Academic lab
Brno, Czechia

BUT Speech@FIT

Brno University of Technology's speech group, author of the VBx and DiariZen diarization toolkits and Czech language models.

speech processingspeaker diarizationCzech LLMs
Independent company
San Francisco, United States

Canopy Labs

A speech AI lab behind Orpheus, a family of Llama-based text-to-speech models released in March 2025 for open use.

text-to-speechspeechopen models
Academic lab
Cardiff, United Kingdom

Cardiff NLP

Cardiff University's NLP group, known for TweetEval, TweetNLP and the Twitter-RoBERTa models for social media analysis.

social media NLP
Independent company
San Francisco, United States

Cartesia

A San Francisco voice AI company founded by the Stanford authors of Mamba-style state-space models, which also releases open Rene and Llamba models.

state-space modelsspeech
Nonprofit
San Francisco, United States

Center for AI Safety

A San Francisco nonprofit reducing societal-scale AI risk through research, field-building and open benchmarks such as HarmBench and WMDP.

safetyevaluationdatasets
Academic lab
Aarhus, Denmark

Center for Humanities Computing

Aarhus University centre applying NLP to the humanities, behind DaCy, odyCy and the Scandinavian Embedding Benchmark.

Danish NLPembeddingsdigital humanities
Independent company
San Francisco, United States

Chai Discovery

A drug-discovery AI startup that released the open Chai-1 biomolecular structure model.

biologystructure prediction
Academic lab
Munich, Germany

CIS, LMU Munich

LMU Munich's language processing centre, behind SimAlign, Glot500 and GlotLID for massively multilingual NLP.

multilingual NLPlow-resource languageslanguage identification
Academic lab
Ljubljana, Slovenia

CJVT

The University of Ljubljana centre that builds Slovene language resources, corpora and the open GaMS generative models.

Slovenelanguage modelsdatasets
Academic lab
Taipei, Taiwan

CKIP Lab

An Academia Sinica research group, active since 1986, that builds Chinese language processing resources and publishes open Chinese NLP models.

Chinese NLPlanguage models
Academic lab
Wroclaw, Poland

CLARIN-PL

Polish language technology infrastructure at Wrocław University of Science and Technology, publishing Polish NLP tools, corpora and benchmarks.

Polish NLPbenchmarks
Institute
Ljubljana, Slovenia

CLASSLA

CLARIN knowledge centre providing open tools, models and corpora for Slovenian, Croatian, Serbian, Bosnian and Montenegrin.

South Slavic NLPlanguage modelscorpora
Academic lab
Antwerp, Belgium

CLiPS

University of Antwerp computational linguistics group, creator of the Pattern web mining library and Dutch NLP resources.

Dutch NLPembeddingsdatasets
Academic lab
Amsterdam, Netherlands

CLTL

VU Amsterdam's Computational Linguistics and Text Mining Lab, publishing open Dutch language resources, MedRoBERTa.nl and teaching material.

Dutch NLP
Nonprofit
Beverly Hills, United States

Common Crawl Foundation

A nonprofit that has published free, open web crawl archives since 2008, forming a foundation dataset for language model research.

datasetsweb crawling
Academic lab
Munich, Germany

CompVis

The Computer Vision and Learning group at LMU Munich, which developed latent diffusion, the research behind Stable Diffusion, and releases its code openly.

visiongenerative modelsdiffusion
Independent company
Mountain View, United States

Contextual AI

An enterprise retrieval-augmented generation company whose research arm releases open alignment methods and retrieval models.

RAGalignmentembeddings
Independent company
Kampala, Uganda

Crane AI Labs

Ugandan AI group building open Luganda and Swahili language models, speech models and datasets on Gemma.

language modelslow-resource languagesdatasets
Community collective
Aarhus, Denmark

Danish Foundation Models

A consortium of Danish universities building open language models, datasets and evaluations so that Danish is not left behind in AI.

language modelsDanishdatasets
Independent company
San Francisco, United States

Deep Cogito

A San Francisco lab releasing open-weight reasoning models trained by iterated distillation and amplification, with the stated aim of general superintelligence.

language modelspost-trainingreasoning
Nonprofit
London, United Kingdom

Deep Learning Indaba

A pan-African machine learning movement that strengthens African AI through an annual Indaba, local IndabaX events, mentorship and awards.

educationcommunitytooling
Independent company
Hangzhou, China

DeepSeek

A Hangzhou AI lab founded by Liang Wenfeng and owned by the hedge fund High-Flyer, known for open-weight V3, R1 and V4 models released under the MIT License.

language modelsreasoningChinese
Independent company
Berlin, Germany

deepset

A Berlin company that created and maintains Haystack, an open-source framework for building LLM and retrieval-augmented applications.

RAGtoolingGerman
Institute
Kaiserslautern, Germany

DFKI

DFKI is Germany's national AI research center, a non-profit public-private partnership founded in 1988 that turns academic research into industrial applications.

AI research
Nonprofit
Jerusalem, Israel

DICTA: Israel Center for Text Analysis

A non-profit center building Hebrew language models and text tools, including the open DictaLM collection.

language modelsHebrewNLP tools
Nonprofit
Kigali, Rwanda

Digital Umuganda

A Rwanda-based team building voice infrastructure and open datasets for African languages, with a focus on Kinyarwanda.

speechdatasetslow-resource languages
Academic lab
Leuven, Belgium

DTAI

The DTAI group at KU Leuven combines machine learning, logic and data mining and is the home of the RobBERT Dutch language models.

Dutch NLPmachine learning
Academic lab
Edinburgh, United Kingdom

EdinburghNLP

EdinburghNLP is the University of Edinburgh's natural language processing group, ranked among the most productive in the world and the home of Nematus.

NLPmachine translationsummarization
Nonprofit
Alexandria, United States

EleutherAI

EleutherAI is a non-profit research institute that grew from a Discord server into a leading source of open language models, datasets and evaluation tools.

language modelsevaluationinterpretability
Independent company
Germany

ellamind

A German AI company, based in Bremen and Berlin, that builds evaluation tools and open-source infrastructure for language models.

evaluationdatasetstooling
Nonprofit
Alicante, Spain

ELLIS Alicante

ELLIS Alicante is a non-profit unit of the European ELLIS network focused on ethical, human-centric AI, directed by Nuria Oliver.

responsible AI
Academic lab
Lausanne, Switzerland

EPFL LLM Team

The EPFL LLM team built Meditron, an open suite of medical language models, and the Megatron-LLM trainer used to produce them.

medical LLMstraining infrastructure
Nonprofit
San Francisco, United States

Epoch AI

A research organization that publishes open datasets and analysis on AI compute, hardware, investment and data centers.

AI trendsdatasetsforecasting
Independent company
San Francisco, United States

Essential AI

A San Francisco lab led by Ashish Vaswani that builds an open platform for AI and has released the Rnj-1 model and the Essential-Web dataset.

language modelsdatasets
Community collective
Lisbon, Portugal

EuroLLM

A series of open multilingual language models covering the 24 official EU languages, built by a European research and industry consortium.

language modelsmultilingualEuropean languages
Independent company
Berlin, Germany

Explosion

A Berlin software company, founded by Matthew Honnibal and Ines Montani, that created the spaCy open-source NLP library and the Prodigy annotation tool.

NLPtoolingmachine learning libraries
Nonprofit
Berkeley, United States

FAR.AI

A nonprofit AI security research organization that builds safety solutions, tests frontier models and convenes the field.

AI safetyadversarial robustnessinterpretability
Institute
San Francisco, United States

fast.ai

A nonprofit research and education group, founded by Jeremy Howard and Rachel Thomas, behind the fastai library and free deep learning courses.

educationlibrariesdeep learning
Institute
Trento, Italy

FBK Machine Translation Unit

The machine translation group at Fondazione Bruno Kessler, publishing open speech and text translation datasets and software.

speech translationspeechdatasets
Institute
Bordeaux, France

Flowers

Flowers is an Inria team in Bordeaux studying curiosity-driven learning in humans and machines, led by Pierre-Yves Oudeyer.

reinforcement learningopen-ended learningLLM agents
Nonprofit
San Francisco, United States

FutureHouse

A nonprofit AI-for-science lab, founded by Sam Rodriques and Andrew White, building language agents to automate scientific discovery.

AI for scienceagentschemistry
Independent company
San Francisco, United States

Genmo

A video generation company that released Mochi 1, an open-source video model under the Apache 2.0 license.

video generation
Independent company
London, United Kingdom

Gensyn

A London-based project building decentralised infrastructure for machine learning, founded by Ben Fielding and Harry Grieve.

decentralised trainingreinforcement learninginfrastructure
Community collective
Accra, Ghana

Ghana NLP

An open-source community building language technology for Ghanaian languages such as Twi, Ewe and Frafra.

low-resource languageslanguage modelsdatasets
Independent company
Paris, France

Giskard

A Paris company building AI red-teaming and evaluation tools, with an open-source testing library and the Phare benchmark.

evaluationsafetyred-teaming
Independent company
San Francisco, United States

Goodfire

An interpretability research company, founded by researchers from OpenAI and Google DeepMind, that works to understand and align AI models.

interpretabilitysafety
Academic lab
Berkeley, United States

Gorilla

A UC Berkeley project that built Gorilla LLM and the Berkeley Function-Calling Leaderboard for evaluating tool and API use by language models.

function callingtool useevaluation
Nonprofit
The Hague, Netherlands

GPT-NL

GPT-NL is a Dutch non-profit effort by TNO, NFI and SURF to build a lawful, sovereign Dutch language model trained on licensed data.

Dutch LLMdatasets
Academic lab
Groningen, Netherlands

GroNLP

GroNLP is the computational linguistics group at the University of Groningen, known for BERTje, HateBERT and Dutch GPT-2 models.

Dutch NLPlanguage models
Independent company
Paris, France

H Company

A Paris AI company building computer-use and web agents and releasing its Holo vision-language models as open weights.

agentscomputer usevision-language models
Academic lab
Stanford, United States

Hazy Research

A Stanford research group led by Chris Re, known for efficient architectures and GPU kernels such as ThunderKittens, Hyena, Based and Monarch Mixer.

architectureskernelsefficiency
Academic lab
Heidelberg, Germany

Heidelberg NLP

Heidelberg NLP is the computational linguistics group at Heidelberg University, producing benchmarks such as VALSE and graph-based language models.

NLP
Academic lab
Helsinki, Finland

Helsinki-NLP

The language technology group at the University of Helsinki, creators of the OPUS-MT open translation models and OPUS parallel corpora.

machine translationdatasetsmultilingual
Academic lab
Donostia-San Sebastian, Spain

HiTZ Basque Center for Language Technology

The Basque Center for Language Technology at the University of the Basque Country, publishing Basque and multilingual language models and NLP tools.

Basquelanguage modelsinformation extraction
Nonprofit
Tehran, Iran

Hooshvare Research Lab

A small non-profit Persian NLP lab that published ParsBERT, the first widely used monolingual BERT for Persian, along with Persian GPT-2 and NER models.

PersianNLP
Independent company
Singapore, Singapore

HPC-AI Tech

Singapore company behind Colossal-AI, an open-source stack for large-model training, and the Open-Sora video generation models.

training infrastructurevideo generation
Institute
Prague, Czechia

HPLT

HPLT is an EU-funded consortium coordinated by Charles University that publishes petabyte-scale multilingual web corpora and models.

multilingual datasetsmachine translationlanguage models
Independent company
Paris, France

Hugging Face

The company behind the Hugging Face Hub and the Transformers, Diffusers and Datasets libraries, the central platform for sharing open models and datasets.

model hublibrariesdatasets
Institute
Budapest, Hungary

Hungarian Research Centre for Linguistics

The Hungarian Research Centre for Linguistics' language technology group publishes the PULI Hungarian language models and national corpora.

Hungarian NLPlanguage modelscorpora
Institute
Martigny, Switzerland

Idiap Research Institute

Independent Swiss non-profit research institute in Martigny working on speech, vision and machine learning, and the home of the original Torch library.

speechvisionmachine learning
Independent company
Paris, France

Illuin Technology

Illuin Technology is a Paris AI company whose research team co-created ColPali visual document retrieval and the ViDoRe benchmark.

document retrievalvision-language modelsNLP
Institute
Athens, Greece

ILSP / Athena Research Center

Greek language technology institute at Athena Research Center that released Meltemi, the first open large language model for Greek, and Llama-Krikri.

Greeklanguage modelsdatasets
Institute
Sofia, Bulgaria

INSAIT

Bulgarian AI and computer science institute at Sofia University, in partnership with ETH Zurich and EPFL, behind the BgGPT and MamayLM open language models.

language modelsBulgarianUkrainian
Academic lab
Linz, Austria

Institute for Machine Learning, JKU Linz

Johannes Kepler University's machine learning institute, led by Sepp Hochreiter, the inventor of LSTM and a driver of modern Hopfield networks and xLSTM.

architecturessciencereinforcement learning
Institute
Warsaw, Poland

Institute of Computer Science PAS

IPI PAN, the Institute of Computer Science of the Polish Academy of Sciences, publishes open Polish retrieval models and question answering datasets.

Polish NLPretrieval
Institute
Abu Dhabi, United Arab Emirates

Institute of Foundation Models

MBZUAI research institute that sponsors LLM360, the fully open effort behind the Amber, K2 and TxT360 releases, and the K2 Horizon model family.

language modelstraining transparency
Institute
Astana, Kazakhstan

ISSAI

Nazarbayev University's institute for AI in Kazakhstan, publishing open Kazakh speech corpora, TTS data and the KazLLM language model.

Kazakhspeechlanguage models
Nonprofit
Belgrade, Serbia

JeRTeh

JeRTeh is a Serbian society that releases Serbian language models, corpora and tools for language technology.

Serbian NLPlanguage modelscorpora
Institute
Stockholm, Sweden

KBLab

Data lab inside the National Library of Sweden that trains open Swedish language and speech models, including KB-BERT and KB-Whisper, on the library's collections.

Swedishspeechlanguage models
Institute
Bratislava, Slovakia

Kempelen Institute of Intelligent Technologies

Independent Slovak non-profit research institute for intelligent technologies, known for SlovakBERT and open Slovak NLP benchmarks.

Slovaklanguage modelsspeech
Independent company
Tokyo, Japan

Kotoba Technologies

Voice AI company that released the Kotoba-Whisper Japanese speech recognition models and builds real-time speech and translation models.

speechJapanese
Nonprofit
Paris, France

Kyutai

Paris non-profit open-science lab building speech-native and multimodal models, creator of Moshi, Mimi, Unmute and Helium.

speechmultimodallanguage models
Nonprofit
Hamburg, Germany

LAION

German non-profit that releases large open datasets, tools and models for AI research, best known for LAION-400M, LAION-5B and OpenCLIP.

datasetsvisionspeech
Community collective
Kyiv, Ukraine

lang-uk

Volunteer community of Ukrainian NLP specialists building open corpora, models and tools for the Ukrainian language.

UkrainianNLPdatasets
Academic lab
Reykjavik, Iceland

Language and Voice Lab

The Language and Voice Lab at Reykjavik University builds open Icelandic speech corpora and recognition and synthesis models.

Icelandicspeech recognitiondatasets
Academic lab
Bangor, United Kingdom

Language Technologies Unit

The Language Technologies Unit at Bangor University builds open Welsh speech, translation and language tools.

Welshspeech recognitionmachine translation
Academic lab
Oslo, Norway

Language Technology Group, University of Oslo

University of Oslo research group in natural language processing behind the NorBERT models, the NorBench benchmark and Norwegian generative models.

Norwegianlanguage modelsbenchmarks
Institute
Santiago, Chile

Latam-GPT

Latin American open language model project coordinated by Chile's national AI centre CENIA, released in 2026 with a regional text corpus.

language modelsdatasetsevaluation
Independent company
Johannesburg, South Africa

Lelapa AI

South African language AI company building resource-efficient models for African languages, maker of Vulavula and the open InkubaLM.

language modelsspeechlow-resource languages
Independent company
San Francisco, United States

Letta

San Francisco AI lab from the creators of MemGPT, building open-source stateful agents that remember and learn over time.

agentsmemory
Independent company
Tokyo, Japan

Lightblue

Tokyo AI company that publishes open Japanese and multilingual language models on Hugging Face alongside its enterprise AI platform.

language modelsrerankingJapanese
Independent company
Zurich, Switzerland

Lightly

Lightly is a Zurich company, born from ETH research, that open-sources self-supervised learning and vision training libraries.

computer visionself-supervised learning
Independent company
New York, United States

Lightning AI

Company behind PyTorch Lightning, LitGPT and LitServe, open-source frameworks for training and serving models.

training infrastructurelibrarieslanguage models
Independent company
Paris, France

LightOn

Paris-based, publicly listed AI company that open-sources retrieval models, ModernBERT variants and multi-vector search tools.

retrievalOCRembeddings
Independent company
Paris, France

LINAGORA

LINAGORA is a French open-source software company behind the LinTO voice assistant and the Lucie French open LLM.

speech recognitionLLMsopen-source software
Independent company
Cambridge, United States

Liquid AI

MIT CSAIL spin-out building efficient on-device foundation models, the LFM family, with open weights on Hugging Face.

language modelsedge AIaudio
Community collective
Tokyo, Japan

LLM-jp

Community of researchers organized by Japan's National Institute of Informatics, building open Japanese-capable large language models.

language modelsJapanesedatasets
Independent company
San Francisco, United States

LMArena

Berkeley-origin team behind Chatbot Arena, the crowd-sourced AI model leaderboard, now operating as the company Arena.

evaluationlanguage modelsdatasets
Independent company
Melbourne, Australia

Maincode

Melbourne AI company that owns its own GPU cluster, builds the Matilda assistant and has released open Maincoder models.

language modelscode
Academic lab
Munich, Germany

MaiNLP

MaiNLP is the Munich AI and NLP lab at LMU Munich, led by Barbara Plank, working on robust and inclusive human-facing NLP.

NLP
Academic lab
Stanford, United States

Marin

Open lab begun at Stanford that develops foundation models in public, releasing code, data, weights and experiment logs for Marin 8B and 32B.

language modelsreproducibility
Academic lab
Abu Dhabi, United Arab Emirates

MBZUAI

Graduate-level AI university in Abu Dhabi that runs the Institute of Foundation Models and publishes open models such as Jais and K2.

researchlanguage modelsvision
Institute
Munich, Germany

MDZ Digital Library team

The MDZ Digital Library team at the Bavarian State Library publishes the widely used dbmdz German, Italian and Turkish BERT models.

language modelsdigital librariesOCR
Independent company
Kuala Lumpur, Malaysia

Mesolitica

Malaysian AI lab building open Malay language models, speech systems and datasets, including MaLLaM and Malaysian Mistral.

Malayspeechtranslation
Nonprofit
Berkeley, United States

METR

A Berkeley nonprofit that measures what frontier AI agents can do on long, autonomous tasks and publishes the evidence for governments, labs and the public.

evaluationsafetyagents
Institute
Montreal, Canada

Mila - Quebec AI Institute

Montreal's AI research institute, founded around Yoshua Bengio's lab, whose members built Theano, SpeechBrain and other widely used open-source tools.

deep learningreinforcement learningresearch
Independent company
Shanghai, China

MiniMax

A Shanghai AI company building multimodal foundation models, with the MiniMax-M series of language models released as open weights on Hugging Face.

language modelsmultimodalimage generation
Independent company
Paris, France

Mistral AI

A Paris AI company that pairs open-weight models under permissive licenses with a commercial platform, positioning itself as Europe's frontier lab.

language modelsmultimodalspeech
Independent company
Berlin, Germany

Mixedbread

Applied research lab building retrieval and reranking models, with open embeddings downloaded more than 50 million times.

embeddingsretrievalrerankers
Independent company
Reykjavik, Iceland

Miðeind

Reykjavik language-technology company building open-source tools, models and benchmarks so that Icelandic is supported in modern AI.

IcelandicNLPlanguage models
Nonprofit
San Francisco, United States

MLCommons

A vendor-neutral consortium that builds and publishes the MLPerf benchmarks, open datasets and AI safety measurements used across the industry.

evaluationdatasetsbenchmarks
Academic lab
Msida, Malta

MLRS

The University of Malta project that builds open corpora, tools and language models for Maltese, a language with few digital resources.

Maltese NLP
Independent company
San Francisco, United States

Moondream

Compact open-weight vision-language models for detection, captioning and segmentation, now part of Prime Intellect.

vision-language modelsedge inferencecomputer vision
Independent company
Beijing, China

Moonshot AI

Beijing lab behind the Kimi models, whose open-weight Kimi K3 family at 2.8 trillion parameters is the largest open release to date.

language modelsmultimodalagents
Independent company
Seoul, South Korea

Motif Technologies

Korean AI lab, a spin-off of the infrastructure firm Moreh, publishing open-weight Motif language models and the training techniques behind them.

language modelstraining efficiencyKorean
Independent company
San Francisco, United States

Mozilla.ai

Mozilla-backed company building open-source tools that give people and organizations choice, transparency and control over how AI is used.

toolingagentsevaluation
Academic lab
Barcelona, Spain

Music Technology Group

The Barcelona research group behind Essentia and Freesound-linked work, building open audio and music analysis software since 1994.

music AIaudio analysis
Institute
Mo i Rana, Norway

NB AI-Lab

The National Library of Norway's AI lab, which releases Norwegian language models, speech models and datasets built from the library's collections.

Norwegianspeechlanguage models
Independent company
London, United Kingdom

Neuphonic

London voice-AI company building low-latency text-to-speech and publishing open on-device NeuTTS models and the NeuCodec audio codec.

speechtext-to-speechaudio codecs
Nonprofit
San Francisco, United States

Neuronpedia

An open-source interpretability platform for exploring, visualizing and steering model internals such as sparse autoencoder features.

interpretabilitysparse autoencoderstooling
Academic lab
Lisbon, Portugal

NLX Group

The University of Lisbon language technology group that released the Albertina and Gervasio open models for Portuguese.

Portuguese NLPlanguage models
Independent company
New York, United States

Nomic AI

New York company that began with open embeddings and the GPT4All local-LLM app and now builds AI agents for architecture, engineering and construction.

embeddingslocal inferencedata visualisation
Institute
Trondheim, Norway

NorwAI

NTNU's national AI innovation centre, publishing Norwegian large language models for research and experimentation.

Norwegian LLMs
Independent company
New York, United States

Nous Research

American open AI lab behind the Hermes models, the Hermes Agent harness and decentralized training work such as DisTrO and Psyche.

language modelsdatasetsagents
Academic lab
Berkeley, United States

NovaSky

Berkeley Sky Computing Lab group releasing open reasoning models, training recipes and the SkyRL reinforcement-learning library for LLM agents.

reasoningreinforcement learningdatasets
Large company
Santa Clara, United States

NVIDIA

A GPU maker that also releases open models, datasets and tooling, from Nemotron language models to Cosmos world models and speech ASR.

language modelsdatasetsspeech
Independent company
Linz, Austria

NXAI

Linz company commercializing Sepp Hochreiter's xLSTM research, publishing xLSTM models, TiRex time-series forecasters and fast recurrent kernels.

xLSTMtime seriesarchitectures
Academic lab
Oxford, United Kingdom

OATML

Yarin Gal's Oxford group on Bayesian deep learning, uncertainty and safe machine learning, with open research code.

uncertaintyBayesian deep learning
Community collective
Germany

Occiglot

An informal European research collective that released Occiglot-7B models, multilingual web datasets and tokenizer analyses for European languages.

language modelsEuropean languagesdatasets
Nonprofit
London, United Kingdom

Open Climate Fix

UK organization applying machine learning to solar, wind and demand forecasting for the grid, publishing its code and data openly.

scienceclimateforecasting
Community collective
Beijing, China

OpenBMB / ModelBest

Open lab for big model base, linked to the Tsinghua NLP group's spin-out ModelBest, releasing the MiniCPM on-device models, VoxCPM speech and UltraRAG.

language modelsmultimodalefficient models
Independent company
Beijing, China

OpenCSG

Initiative founded in 2023 that builds the CSGHub platform for managing models, datasets and agents, and releases open Chinese datasets and models.

datasetslanguage modelsChinese
Community collective
Prague, Czechia

OpenEuroLLM

EU-funded consortium of universities, companies and supercomputing centres building open, multilingual foundation models for the languages of Europe.

language modelsmultilingualEuropean languages
Community collective
Sankt Augustin, Germany

OpenGPT-X

German publicly funded consortium that trained the open Teuken-7B models, built from scratch for all 24 official languages of the European Union.

language modelsmultilingualEuropean languages
Independent company
Boston, United States

OpenHands

Open-source platform for AI software-engineering agents, built by the former All Hands AI team and used by hundreds of contributors.

coding agentssoftware engineeringagents
Community collective
Paris, France

OpenLLM France

French community and BPI-funded consortium building open commons for French, including the Lucie-7B and Luciole model families.

language modelsFrenchdatasets
Community collective
Bucharest, Romania

OpenLLM-Ro

Community of Romanian researchers releasing open Romanian language models, datasets, benchmarks and training code.

language modelsRomaniandatasets
Academic lab
Shanghai, China

OpenMOSS

Fudan-linked research group at the Shanghai Innovation Institute behind the MOSS models, MOSS-TTS speech and image releases and interpretability tools.

language modelsspeechimage generation
Academic lab
Mountain View, United States

OpenThoughts

A research collective curating fully open reasoning datasets and recipes, from the OpenThoughts3 data recipe to the OpenThinker models.

reasoning datadatasetspost-training
Independent company
Seattle, United States

Oumi

A public benefit company building an unconditionally open platform for training, evaluating and deploying foundation models, with an open-source library at its core.

training platformevaluationhallucination detection
Independent company
Tehran, Iran

Part AI

A Tehran-based developer of Persian-language models, known for the Dorna decoder LLMs and the TookaBERT and Tooka-SBERT encoders.

Persianlanguage modelsembeddings
Independent company
Paris, France

Photoroom

The Paris image-editing company that trains its own text-to-image models and releases them as open source under Apache 2.0.

image generationdiffusioncomputer vision
Independent company
San Francisco, United States

Physical Intelligence

A robotics foundation-model company whose pi0 family of vision-language-action models and openpi code give robot learning a shared open baseline.

roboticsvision-language-action modelsfoundation models
Independent company
Paris, France

Pleias

A Paris research company building language models and the Common Corpus dataset on openly licensed and rights-cleared data.

datasetslanguage modelsopen data
Independent company
San Francisco, United States

Pluralis Research

A research lab developing Protocol Learning, where large models are trained across many untrusted participants without any one holding the full weights.

distributedtraining
Academic lab
Bonn, Germany

Polygl0t

A University of Bonn project building open foundation models for low-resource languages, best known for the Tucano Portuguese models and the GigaVerbo corpus.

language modelslow-resource languagesdatasets
Academic lab
New York, United States

Polymathic AI

A Simons Foundation and Flatiron Institute backed collaboration building open foundation models and datasets for scientific data.

scientific foundation modelsphysics simulationastronomy
Independent company
Tokyo, Japan

Preferred Networks

A Tokyo deep learning company behind Chainer, the MN-Core accelerator and the PLaMo Japanese language models.

language modelsJapanesefinance
Independent company
San Francisco, United States

Prime Intellect

A lab and compute platform working on distributed training, asynchronous reinforcement learning and open environments for agentic models.

distributedlanguage modelsreinforcement learning
Academic lab
Princeton, United States

Princeton NLP Group

The Princeton NLP group, source of SWE-bench, SimPO and Tree of Thoughts, with many open benchmarks, methods and models.

benchmarksalignmentagents
Independent company
Freiburg, Germany

Prior Labs

A Freiburg company behind TabPFN, the tabular foundation model published in Nature that predicts on structured data without task-specific training.

tabular datafoundation modelstime series
Independent company
Paris, France

probabl

A Paris mission-driven company spun out of Inria to sustain scikit-learn and the open-source data science ecosystem around it.

tabular datatoolingevaluation
Institute
Barcelona, Spain

Projecte Aina

A Catalan government initiative led by the Barcelona Supercomputing Center that builds open Catalan language models, speech resources and corpora.

Catalanlanguage modelsspeech
Academic lab
Santiago de Compostela, Spain

Proxecto Nos

The Xunta de Galicia project, run by USC, that creates open Galician language models, corpora and speech tools.

Galicianlanguage modelsspeech
Independent company
Paris, France

Pruna AI

A model efficiency company that open-sources the pruna framework for compressing and accelerating AI models.

efficiencyquantisationimage generation
Independent company
Paris, France

pyannoteAI

The creators of pyannote.audio, the open-source speaker diarization toolkit, now also offering hosted models through pyannoteAI.

speechspeaker diarizationtooling
Institute
Doha, Qatar

Qatar Computing Research Institute

Qatar's national computing research institute at HBKU, developer of Arabic language technology including the Fanar generative AI platform.

ArabicNLPdatasets
Independent company
Sunnyvale, United States

Reka

A multimodal AI company that releases selected open-weight models such as Reka Flash 3 and Reka Flash 3.1.

multimodallanguage modelsevaluation
Independent company
Tokyo, Japan

rinna

A Japanese generative AI company known for releasing pretrained Japanese GPT, BERT and speech models, including the Youri and Nekomata LLMs.

language modelsJapanesespeech
Independent company
Des Moines, United States

Roboflow

A computer vision platform company that open-sources Supervision, Inference, RF-DETR and the Universe dataset hub.

computer visionobject detectiondatasets
Community collective
Hong Kong, Hong Kong

RWKV

An open-source community behind the RWKV recurrent language model architecture, hosted as a Linux Foundation AI & Data project.

language modelsarchitecturetraining infrastructure
Independent company
Tokyo, Japan

Sakana AI

A Tokyo research company founded by former Google researchers, known for evolutionary model merging and The AI Scientist.

language modelsevolutionary methodsdatasets
Academic lab
Rome, Italy

Sapienza NLP

The Sapienza University of Rome NLP group led by Roberto Navigli, creator of BabelNet and the Minerva Italian language models.

language modelsItaliansemantics
Independent company
Bengaluru, India

Sarvam AI

An Indian AI company building open and sovereign language models for Indian languages, including Sarvam 30B and 105B.

language modelstranslationIndic languages
Independent company
San Francisco, United States

Sesame

A conversational speech company that open-sourced the CSM-1B base speech model while building voice companions.

speech
Institute
Shanghai, China

Shanghai AI Laboratory

A state-backed Chinese research laboratory behind the open Intern family of language and multimodal models: InternLM, InternVL and Intern-S1.

language modelsmultimodaltoolchain
Nonprofit
Barcelona, Spain

Softcatala

A volunteer-driven Catalan non-profit that makes software, translation and speech tools available in Catalan.

Catalanmachine translationspeech
Community collective
Warsaw, Poland

SpeakLeash

A Polish open community collecting Polish text and building the Bielik models with ACK Cyfronet AGH.

language modelsPolishdatasets
Academic lab
Gothenburg, Sweden

Sprakbanken Text

The University of Gothenburg's national language data infrastructure, building Korp, Sparv and open resources for Swedish since 1975.

Swedish NLPcorpora
Academic lab
Zurich, Switzerland

SRI Lab

ETH Zurich's group on trustworthy machine learning, producing neural network verifiers, LMQL and ML-for-code tools.

LLM programmingrobustness
Independent company
London, United Kingdom

Stability AI

A London generative AI company behind Stable Diffusion that publishes open-weight image, video, audio and language models, many under community licences.

image generationvideoaudio
Academic lab
Stanford, United States

Stanford CRFM

Stanford's Center for Research on Foundation Models, creators of the HELM evaluation framework and Alpaca.

evaluationlanguage modelspolicy
Academic lab
Stanford, United States

Stanford NLP Group

The Stanford NLP Group is a leading academic research group whose open software, including CoreNLP, GloVe, Stanza and DSPy, underpins much of modern language technology.

NLPLM programmingdatasets
Independent company
Shanghai, China

StepFun

StepFun is a Shanghai foundation-model company that has released open-weight language, speech and video models under permissive licenses.

language modelsmultimodalspeech
Independent company
Tokyo, Japan

Stockmark

Stockmark is a Tokyo company that has released large Japanese-language models, including the 100-billion-parameter Stockmark-100b, under permissive licenses.

language modelsdocumentsJapanese
Academic lab
Tokyo, Japan

Swallow LLM

Swallow is a Japanese LLM project at the Institute of Science Tokyo that continually pretrains open models to give them strong Japanese ability.

language modelsJapanesemedical
Academic lab
Bari, Italy

SWAP Research Group

Bari's SWAP group, creators of the LLaMAntino family of open Italian language models.

Italian LLMsrecommender systems
Institute
Zurich, Switzerland

Swiss AI Initiative

The Swiss AI Initiative is a national ETH Zurich and EPFL effort that trained the fully open Apertus language models on the Alps supercomputer.

language modelsmultilingualdatasets
Institute
Taipei, Taiwan

TAIDE

TAIDE is a Taiwanese government-backed project releasing open Traditional Chinese language models for trustworthy local use.

language modelsTraditional Chinese
Academic lab
Zagreb, Croatia

TakeLab

Zagreb's text analysis and knowledge engineering lab, a leading Croatian NLP group.

NLPCroatiandata loading
Academic lab
Tallinn, Estonia

TalTech Language Technology

The TalTech language technology lab publishing open Estonian speech recognition and translation models.

Estonianspeech recognitionlanguage identification
Academic lab
Tartu, Estonia

TartuNLP

TartuNLP is the University of Tartu's language technology group, building Estonian and Finno-Ugric models, translation and speech tools.

Estonianspeechmachine translation
Nonprofit
Kaitaia, New Zealand

Te Hiku Media

Te Hiku Media is a Māori iwi media organization in Kaitaia that builds te reo Māori speech technology under Māori data sovereignty.

speechdata sovereigntylow-resource languages
Institute
Abu Dhabi, United Arab Emirates

Technology Innovation Institute

TII is Abu Dhabi's government-funded research institute and the publisher of the Falcon open language models.

language modelsvision1-bit models
Institute
London, United Kingdom

The Alan Turing Institute

The Alan Turing Institute is the UK's national institute for data science and AI, which maintains a large body of open research software.

research softwaredatasetsdata science
Independent company
Riga, Latvia

Tilde

Tilde is a Latvian language technology company that trained TildeOpen, an open 30B model covering 34 European languages.

language modelsmachine translationBaltic languages
Independent company
Munich, Germany

TNG Technology Consulting

TNG is a German technology consultancy that publishes open-weight DeepSeek merges such as the R1T Chimera models.

language modelsmodel merging
Independent company
San Francisco, United States

Together AI

Together AI is an AI cloud company whose research arm released RedPajama and advanced open work such as FlashAttention-2 and Mixture-of-Agents.

datasetslanguage modelstraining infrastructure
Nonprofit
San Francisco, United States

Transluce

Transluce is a nonprofit AI research lab building open tools to understand and oversee AI systems.

interpretabilityevaluationsafety
Independent company
Seoul, South Korea

Trillion Labs

Trillion Labs is a Seoul startup that trains Korean and English foundation models from scratch and releases open weights.

language modelsKoreanOCR
Independent company
Kashiwa, Japan

Turing Inc.

Turing is a Tokyo autonomous-driving company that releases the Heron vision-language models and datasets.

vision-languageautonomous driving
Academic lab
Turku, Finland

TurkuNLP

TurkuNLP is the University of Turku's language technology group, known for FinBERT and, with Silo AI, the Poro and Viking models.

language modelsFinnishdatasets
Academic lab
Darmstadt, Germany

UKP Lab

The Ubiquitous Knowledge Processing Lab at TU Darmstadt created Sentence-BERT and publishes open NLP tools.

NLPembeddingsdatasets
Independent company
Frederick, United States

Ultralytics

Ultralytics develops the YOLO model family and ultralytics library, released under AGPL-3.0 with enterprise licensing.

object detectioncomputer visionYOLO
Independent company
San Francisco, United States

Unsloth

Unsloth is an open-source project and company making LLM fine-tuning and reinforcement learning faster and lighter.

fine-tuningtraining infrastructurequantization
Independent company
Seoul, South Korea

Upstage

Upstage is a Korean AI company behind the Solar language models and document processing engines, some released openly.

language modelsdocumentsKorean
Institute
Toronto, Canada

Vector Institute

Vector Institute is a Toronto nonprofit AI research institute that also publishes open tools for LLM work, federated learning and health AI.

toolingfederated learningevaluation
Academic lab
Rayong, Thailand

VISTEC-depa AI Research Institute of Thailand

VISTEC-depa AIRI is a Thai institute publishing Thai NLP datasets and models such as WangchanBERTa and WangchanLION.

Thaidatasetslanguage models
Academic lab
Oxford, United Kingdom

Visual Geometry Group

Oxford's Visual Geometry Group, source of VGG networks and open vision datasets and tools.

computer vision3D visiondatasets
Academic lab
Lausanne, Switzerland

VITA

EPFL's lab on socially aware AI for transportation, with open pose and trajectory code.

visionautonomous driving
Independent company
Ann Arbor, United States

Voxel51

Voxel51 builds the open-source FiftyOne toolkit for curating datasets and evaluating computer vision models.

computer visiondatasetsevaluation tooling
Academic lab
Istanbul, Turkey

YTU COSMOS AI Research Group

Yildiz Technical University's COSMOS group, publishing open Turkish language models.

Turkish LLMs
Independent company
Beijing, China

Z.ai

Z.ai, formerly Zhipu AI, is a Tsinghua spinoff that releases the GLM models, including MIT-licensed open weights.

language modelsagentsevaluation
Independent company
San Francisco, United States

Zyphra

Zyphra is an AI lab in Palo Alto releasing open-weight models, including the ZAYA1-8B reasoning mixture-of-experts model.

speechlanguage modelsstate-space models
Academic lab
Prague, Czechia

ÚFAL

ÚFAL is Charles University's Institute of Formal and Applied Linguistics, behind UDPipe and Czech treebanks.

NLPCzechparsing