EuroLLM-9B: Technical Report

TL;DR

EuroLLM-9B is a large multilingual language model designed specifically for European languages, addressing underrepresentation issues and demonstrating competitive performance on benchmarks, with all components openly available.

Contribution

The paper introduces EuroLLM-9B, a new open-source multilingual LLM tailored for European languages, including novel datasets and filtering techniques to improve language coverage.

Findings

EuroLLM-9B performs well on multilingual benchmarks.

EuroLLM-9B advances European language support in open LLMs.

All components are publicly released for research use.

Abstract

This report presents EuroLLM-9B, a large language model trained from scratch to support the needs of European citizens by covering all 24 official European Union languages and 11 additional languages. EuroLLM addresses the issue of European languages being underrepresented and underserved in existing open large language models. We provide a comprehensive overview of EuroLLM-9B's development, including tokenizer design, architectural specifications, data filtering, and training procedures. We describe the pre-training data collection and filtering pipeline, including the creation of EuroFilter, an AI-based multilingual filter, as well as the design of EuroBlocks-Synthetic, a novel synthetic dataset for post-training that enhances language coverage for European languages. Evaluation results demonstrate EuroLLM-9B's competitive performance on multilingual benchmarks and machine translation tasks, establishing it as the leading open European-made LLM of its size. To support open research and adoption, we release all major components of this work, including the base and instruction-tuned models, the EuroFilter classifier, and the synthetic post-training dataset.