[AI Summary]: EuroLLM is a collaborative multilingual large language model developed by European research institutions, supporting all 24 official EU languages. Built with 9 billion parameters and trained on over 4 trillion tokens across 35 languages, the model aims to advance EU AI sovereignty and technological independence. Two versions are publicly available via Hugging Face: EuroLLM 9B Base for fine-tuning applications, and EuroLLM 9B Instruct for conversational and instruction-following tasks. The model supports multilingual question-answering, summarization, and translation, though its practical performance has received mixed reviews.
- Developer: European Research Consortium
- License: Open source (Hugging Face)
- Platform: Hugging Face
- Model Size: 9B parameters
- Languages: 24 official EU languages (Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, Swedish) + 11 additional languages
- Training Data: 4+ trillion tokens across 35 languages
- Language: Korean (documentation), Multilingual (model)