EuroLLM: LLM made in Europe built to support all 24 official EU languages

Large language model made in Europe built to support all official 24 EU languages

Soon we will be adding vision and voice to our models so that they can interpret and understand images and speech.

Freely used by researchers, organisations and citizens of Europe.

Great on language related tasks, including question answering, summarisation, and translation.

Models pretrained and finetuned on text from all languages.

Our current flagship model. A 9B parameter model trained on over 4 trillion tokens of multilingual data across 35 different languages, including all official EU languages. We’ve made EuroLLM 9B Base available for fine-tuning on any task. As a demonstration, we’ve also provided EuroLLM 9B Instruct, a model fine-tuned for instruction following and chat capabilities.

A 1.7B parameter model trained on similar data to EuroLLM-9B, that is ideal to for use in edge devices.

EuroLLM release Article on Hugging Face LINK

Technical Report for EuroLLM 1.7B LINK

Sharing a common vision, our team is committed to advancing multilingual AI technologies to empower Europe’s digital future and strengthen the EU’s commitment to AI sovereignty. The team’s goal is for EuroLLM to become a flywheel for innovation — offering anyone the opportunity to use this EU homegrown LLM and build upon it. The project is living proof that amazing things can happen when Europe comes together to push the boundaries of innovation.

Source: HackerNews (https://news.ycombinator.com/item?id=45733707)