Evidence-backed European AI register
EuroLLM
Reviewed identity, roles, model listings, pricing and evidence for EuroLLM.
Dataset compiled 2026-08-27 · evidence over inference
Reviewed identity
- Headquarters scope
- EU headquarters
- Headquarters record
- Lisbon
- Legal entity
- no separate legal entity; an academic/industry consortium operating under the Horizon Europe RIA project UTTER (grant 101070631) and an EuroHPC Extreme Scale Access grant
- Ownership
- public research + industry consortium; models published under the HuggingFace org 'utter-project'. Lead scientific institutions: Instituto Superior Técnico / Instituto de Telecomunicações (Universidade de Lisboa, PT) and Unbabel (PT)
- Founded
- 2024
- Confidence
- high
- Last verified
- 2026-08-22
Evidence boundaries
Organisation listings identify a reviewed canonical record. Model listings show an association in the source register; they do not by themselves establish model creation, current API availability, hosting location or infrastructure ownership.
Model listings
- Type
- text
- Parameters
- 22.64B total (21.07B non-embedding), dense transformer, GQA, RMSNorm pre-LN, SwiGLU, RoPE
- Context
- 32,768 tokens
- Languages
- 35 languages: all 24 official EU languages plus Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, Ukrainian (11 additional strategic languages)
- Licence
- Apache-2.0
- Release
- 2026-02
- Type
- text
- Parameters
- 9B dense
- Context
- 4,096 tokens
- Languages
- 35 languages (24 official EU + 11 additional)
- Licence
- Apache-2.0
- Release
- 2024-12 (9B), refreshed 2025-12 as EuroLLM-9B-2512
- Type
- text
- Parameters
- 1.7B dense
- Context
- 4,096 tokens
- Languages
- 35 languages
- Licence
- Apache-2.0
- Release
- 2024-09
- Type
- text
- Parameters
- MoE, 2.6B total / ~0.6B active
- Context
- unknown
- Languages
- EU languages (same target set)
- Licence
- Apache-2.0
- Release
- 2025-12
- Type
- multimodal
- Parameters
- 9B
- Context
- unknown
- Languages
- EU languages
- Licence
- Apache-2.0
- Release
- 2025
- Type
- multimodal
- Parameters
- 3B / 10B
- Context
- unknown
- Languages
- multilingual, translation-centric
- Licence
- unknown
- Release
- 2025
- Type
- video
- Parameters
- 3B / 10B
- Context
- unknown
- Languages
- multilingual
- Licence
- unknown
- Release
- 2025-2026
- Type
- asr
- Parameters
- 94.4M (speech representation / feature extraction)
- Context
- unknown
- Languages
- 147 languages
- Licence
- unknown
- Release
- 2024
- Spire (SpireBase / SpireFull / SpireNoBlocks / SpireNoPseudo)
- Type
- multimodal
- Parameters
- 7B speech-augmented LLM variants
- Context
- unknown
- Languages
- multilingual speech+text
- Licence
- unknown
- Release
- 2025
Profile note
The most concretely productive pan-European open-model consortium to date: 24 official EU languages plus 11 strategic ones, Apache-2.0, and a full family spanning 1.7B/9B/22B dense, a small MoE, a VLM, video models and speech encoders. Members: Unbabel, Instituto Superior Técnico, Instituto de Telecomunicações, University of Edinburgh, Université Paris-Saclay, Sorbonne Université, Naver Labs Europe, University of Amsterdam, and Aveni. Its funding shape is instructive — very little cash, a large EuroHPC compute grant, and an existing Horizon Europe MT project (UTTER) as the legal vehicle. No s…
Record sources
- https://eurollm.io/
- https://huggingface.co/utter-project/EuroLLM-22B-Instruct-2512
- https://huggingface.co/utter-project
- https://huggingface.co/blog/eurollm-team/eurollm-22b
- https://arxiv.org/pdf/2602.05879
- https://tecnico.ulisboa.pt/en/news/eurollm-22b-open-european-artificial-intelligence-model-has-been-launched-…