innFactory AI Consulting from Rosenheim, Germany is closely tracking OpenEuroLLM – the project directly addresses the sovereign-AI and data-protection requirements we work on with DACH-region customers every day. Once the first models are released, we will support evaluations against established US and Chinese alternatives. This overview reflects the state as of October 2026.
What is OpenEuroLLM?
OpenEuroLLM is a pan-European initiative led by 20 research and industry partners with the goal of building a truly open-source LLM for the EU. The project is funded by the EU Digital Europe Programme with around EUR 20 million (of a total project budget of EUR 37.4 million).
Core partners (selection)
- Aleph Alpha (Heidelberg) – German LLM and tokenizer expertise
- Ellamind – model and training know-how
- Prompsit Language Engineering – European language processing
- AMD – hardware stack and compute
- Barcelona Supercomputing Center and other EU HPC centres
- Tübingen AI Center, AI Sweden and other academic partners
Project goals
- Truly open source: open weights, open training data, open training pipelines
- All 24 EU official languages plus accession-candidate languages
- Digital sovereignty: training and inference inside the EU
- Strategic compute allocation on European supercomputers
Roadmap
- Already achieved: Per the project blog post “First year progress and next steps” (Tübingen AI Center, 6 March 2026), the collaboration with the HPLT project has already produced reference models in 2B and 100B parameter sizes for various languages – an intermediate step, not the final OpenEuroLLM model.
- Mid 2026 (target): Initial release version (planned: 8B model)
- Summer 2026: First larger model using strategically allocated compute
- 2028: Final consortium version with full language coverage
We found no announcement on the project website (openeurollm.eu) of a released 8B model; the target set for summer 2026 therefore does not appear to be publicly confirmed as met.
Strategic relevance for DACH enterprises
OpenEuroLLM is relevant for mid-market and public-sector buyers in the DACH region for several reasons:
- EU sovereignty narrative: Unlike Mistral (France) or Aleph Alpha (merger with Cohere agreed, closing pending), OpenEuroLLM is consortium-led and not dependent on a single commercial vendor.
- Open licence: Self-hosting in regulated industries (banking, insurance, government, healthcare) without API dependency.
- Language fairness: Unlike US models that significantly favour English, OpenEuroLLM aims to perform on par for German, French or Italian communication.
- EU AI Act: EU-funded open-source models benefit from simplified compliance pathways.
Current status
The final OpenEuroLLM model has, per our research, still not been released. Published so far are concept and roadmap documents, interim reports (such as “First year progress” from the Tübingen AI Center, March 2026), and reference models (2B/100B) made available through the HPLT collaboration. For production workloads today, we recommend evaluating in parallel:
- Mistral for open, EU-based LLMs at current maturity
- Aleph Alpha (Kolibri-1, Pharia) for German-language focus
- Qwen 3.6 (Apache 2.0) for self-hosting with broad language coverage
Our recommendation
OpenEuroLLM is a strategically relevant EU sovereign-AI project, but operationally still a “watch” item. We recommend that enterprises whose roadmap includes an EU sovereignty strategy actively track the project and prepare pilot setups as soon as first models become available. For production today, Mistral or Aleph Alpha (Kolibri-1, PhariaAI) are the obvious entry points.
Contact us for an assessment of which EU sovereignty strategy fits your organisation.