EuroLLM
Open multilingual models for European languages
EuroLLM is a project developing open-weight language models focused on European Union languages. Released in September 2024, the initial 1.7 billion parameter models aim to address the English-centric bias in existing large language models by supporting all official EU tongues.
- Supports all official European Union languages
- Initial models have 1.7 billion parameters
- Released as open-weight models in 2024
- Includes base and instruction-tuned variants
- Features a custom multilingual tokenizer
What it is
EuroLLM is a research initiative creating open-weight large language models specifically designed for Europe. The project addresses the dominance of English in current AI systems by training models on all official languages of the European Union. The first release includes two variants: a base model and an instruction-tuned version, both with 1.7 billion parameters. These models are intended for general text understanding and generation across multiple European tongues.
How it came about
The project was announced in September 2024 via a research paper on arXiv. The team outlined their progress in data collection, filtering, and the development of scaling laws specific to multilingual contexts. This initial release marks the first public availability of the EuroLLM-1.7B and EuroLLM-1.7B-Instruct models, representing the foundational step in their roadmap for European language support.
Inside the model
The models use a decoder-only transformer architecture. A key component of the design is a custom multilingual tokenizer created to handle the diverse linguistic structures of European languages efficiently. The training process involved careful data mixing and modeling configurations derived from newly developed scaling laws. The team focused on curating and filtering data to ensure high quality across all target languages, rather than relying solely on English-centric datasets.
What it can do
The initial models are evaluated on multilingual general benchmarks and machine translation tasks. They are designed to understand and generate text in all official EU languages, as well as several other relevant regional languages. Performance metrics reported in the initial paper focus on these multilingual capabilities, demonstrating the model's ability to operate effectively outside of English-dominated contexts.
How to use it
The weights for EuroLLM-1.7B and EuroLLM-1.7B-Instruct are released openly. This allows researchers and developers to download, inspect, and fine-tune the models without restriction. The project emphasizes openness as a core value, aiming to facilitate broader access to high-quality multilingual AI tools for the European community and beyond.
Questions people ask about EuroLLM
Which languages does EuroLLM support?
EuroLLM is designed to understand and generate text in all official languages of the European Union. It also includes support for several additional relevant languages beyond the EU official list, aiming for broad European coverage.
Is EuroLLM free to use?
Yes, the project releases open-weight models. This means the model weights are publicly available for download and use, allowing for local deployment and further development without API fees or proprietary restrictions.
How large is the EuroLLM model?
The initial release features models with 1.7 billion parameters. This size classifies it as a smaller model, suitable for efficient deployment while still maintaining capable multilingual performance.
Good reading and sources
There is no English Wikipedia article on EuroLLM yet, so the facts on this page come from the sources listed above: arxiv.org/abs/2409.16235. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.