OpenEuroLLM, the European open‐source project for large language models (LLMs) that aims to serve as the regional alternative to systems like ChatGPT, Gemini, Grok, and other artificial intelligence (AI) projects, has a launch date. The first version is expected to be ready by next year.

Behind the project is a consortium of 20 European organizations, including research centres, high-performance computing facilities, and companies such as AMD, Aleph Alpha, Ellamind, Prompsit Language Engineering, and high-performance computing (HPC) centres such as the Barcelona Supercomputing Center. Although preparations began a year ago, the project was only recently made public, and its leaders anticipate that the initial version will be available by mid-2026. The project is scheduled to conclude two years later, with the final release published on GitHub.

Officially presented in February, the initiative is “aligned with the imperative to improve Europe’s competitiveness and digital sovereignty,” according to the European Commission. Furthermore, it aims to develop a family of open-source large language models that reflect the linguistic diversity of the European Union (EU). The project presents itself as truly open, EU regulations compliant, and linguistically and culturally diverse. The final version of OpenEuroLLM will see the light in 2028, promising to offer access in all EU languages.

A 24-language AI project

The initiative is co-led by computational linguist Jan Hajic from Charles University in Prague (Czech Republic) and Peter Sarlin, the executive director and co-founder of Finnish AI lab Silo AI (affiliated with AMD). OpenEuroLLM positions itself as the regional alternative to international LLMs such as GPT (OpenAI), Llama (Meta), Gemini (Google), Qwen (Alibaba), and DeepSeek. According to the project’s website, OpenEuroLLM multilingual models “preserve both linguistic and cultural diversity, enabling European companies to develop high-quality products and services in the era of AI.”

The model will support the 24 official EU languages, including Albanian, considering that Albania is currently negotiating its entry into the EU. The project has a price tag of 37.4 million euros, with roughly 20 million euros provided by the EU’s Digital Europe Programme. While its budget is comparable to projects like DeepSeek, it remains modest compared to the multi-billion-euro investments of major global players in AI infrastructure.

While the initial funding is relatively modest, the EU has made a significant investment in another infrastructure project, Iris2, to compete with Starlink in satellite manufacturing. The cost of the European satellite constellation was initially set at 6 billion euros but later increased to 11.4 billion euros.

Open source: An ambitious goal

The European initiative to create an open-source alternative to major AI models faces several challenges. Although the project aims to be fully open, it may need to make compromises to meet its quality standards.

Due to European copyright laws and the requirement to ensure high-quality models, some training data might be kept private but accessible to auditors, as mandated by the EU AI Act for high-risk AI systems. While the project intends to make as much data available as possible (especially data from sources like open repositories of web data like Common Crawl), full transparency remains a goal subject to regulatory compliance.

Moreover, maintaining consistent quality across all languages will be challenging, as the LLM must gather a comparable volume of data for each language. Therefore, OpenEuroLLM will face greater linguistic, regulatory, and financial hurdles than other AI projects.