Models published by European providers, models hosted in the EU, and models from elsewhere that European teams can run here. Every card states what is documented and what is not.
This page is a static copy: the URL you see cannot change what it renders, so there is no filter form here. All 53 models are listed below, and the console offers the same catalogue with filters.
Get an API keyModels found: 53
Codestral is Mistral’s code-completion model, served through the Mistral API rather than published as weights. It supports fill-in-the-middle, tool ca…
Devstral Small 2 is a 24B Apache-2.0 model fine-tuned for software-engineering agents, built with All Hands AI. The model card reports 68.0% on SWE-be…
EuroLLM-1.7B-Instruct is the smallest model of the EuroLLM family, trained on 4 trillion tokens across the same 35 languages as the larger sizes. It i…
EuroLLM-22B-Instruct (2512) is the largest released EuroLLM model, covering the same European language set as the 9B version. The model card states th…
EuroLLM-9B-Instruct (2512) is the long-context revision of the EuroLLM 9B instruct model, trained for the 24 official EU languages plus Catalan, Galic…
LLMEU V2 is the model LLM EU intends to train and serve on European infrastructure, aimed at the gap the current generation leaves open: the EU's smal…
Magistral Small 1.2 is a 24B Apache-2.0 reasoning model derived from Mistral Small 3.2, with a vision encoder and explicit think tokens for the reason…
Ministral 3 8B is a small Apache-2.0 instruct model with vision input and a 256k-token context window. Mistral targets edge deployment with it: the ve…
Mistral 7B Instruct v0.3 is the extended-vocabulary version of Mistral’s original 7B instruct model, with a 32k-token context window and function-call…
Mistral Large 3 is the largest published Mistral model: a granular mixture-of-experts model with 675B total parameters and 41B active parameters, rele…
Mistral NeMo is a 12B model built with NVIDIA, released under Apache-2.0 with a 128k-token context window and a Tekken tokenizer. The card reports 68.…
Mistral Small 4 is an Apache-2.0 mixture-of-experts model with 119B total parameters and 6.5B activated per token. It accepts text and images and gene…
This is a placeholder for the final OpenEuroLLM model release, scheduled in the project plan for 31 January 2028 together with the training and evalua…
This is a placeholder for the first OpenEuroLLM model release, which the project lists as a deliverable due on 31 December 2026. No weights, tokenizer…
Pharia-1-Embedding-4608-control is an embedding model that Aleph Alpha built on top of Pharia-1-LLM-7B-control and published under the same Open Aleph…
Pharia-1-LLM-7B-control-aligned is Aleph Alpha’s 7B control model, published on Hugging Face under the Open Aleph License, which limits use to educati…
Pixtral 12B is Mistral’s first vision-language model, released as open weights under Apache-2.0. Its configuration file declares a maximum position em…
Teuken 7B Instruct v0.6 is a German publicly funded model trained on all 24 official EU languages, developed by Fraunhofer, Forschungszentrum Jülich, …
Voxtral Small is an Apache-2.0 audio-input model that transcribes and reasons over speech, with a 32k-token context that the vendor maps to roughly 30…
Aya 23 8B is the 8B member of Cohere Labs’ Aya 23 release, covering 23 languages under CC-BY-NC-4.0 with a context length of 8192 tokens. It is the mu…
Aya Expanse 8B is Cohere Labs’ multilingual instruction model covering 23 languages, released under CC-BY-NC-4.0. Its metadata declares 23 languages, …
Command A+ is Cohere’s 219B-parameter flagship released as open weights under Apache-2.0, with image input and a model card that states 128K input and…
Command R (08-2024) is a 32B retrieval-augmented-generation model with a 128K-token context window, published under CC-BY-NC-4.0. That licence allows …
DeepSeek-R1 is the reasoning model that made open-weight chain-of-thought training widely reproducible; it has 671B total and 37B activated parameters…
These four labels replace the bare word sovereign, which we do not use on its own. A model is called sovereign only with its class and region beside it, because ownership, operation and jurisdiction are three different facts.