Evidence-backed European AI register
TartuNLP — University of Tartu
Reviewed identity, roles, model listings, pricing and evidence for TartuNLP — University of Tartu.
Dataset compiled 2026-08-27 · evidence over inference
Reviewed identity
- Headquarters scope
- EU headquarters
- Headquarters record
- Tartu
- Legal entity
- NLP research group at the Institute of Computer Science, University of Tartu (public university)
- Ownership
- public research
- Founded
- unknown
- 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
- 71B — continued pretraining of Llama-3.1-70B on ~60B tokens, then SFT and preference optimisation
- Context
- unknown
- Languages
- Estonian, English
- Licence
- Llama 3.1 Community License
- Release
- 2024-08-26 (per the 0826 version tag)
- Type
- text
- Parameters
- 71B
- Context
- unknown
- Languages
- Estonian, English
- Licence
- Llama 3.1 Community License
- Release
- 2024-08
- Type
- text
- Parameters
- 8B
- Context
- unknown
- Languages
- Estonian, English
- Licence
- Llama 3.1 Community License
- Release
- 2025
- Apertus-EstLLM-8B-Instruct-1125
- Type
- text
- Parameters
- 8B — Estonian post-training of the Swiss Apertus open model
- Context
- unknown
- Languages
- Estonian, English
- Licence
- unknown
- Release
- 2025-11
- Llammas (Estonian Llama family, incl. GEC variants)
- Type
- text
- Parameters
- 7B class
- Context
- unknown
- Languages
- Estonian
- Licence
- unknown
- Release
- 2024
- EstBERT_NER_v2, est-roberta-* (hist-ner, morph tagging, homonym tagging)
- Type
- text
- Parameters
- 0.1B each
- Context
- 512 tokens
- Languages
- Estonian
- Licence
- unknown
- Release
- unknown
Profile note
The de facto Estonian national LLM programme, delivered by a university group rather than a company: 78 models on HuggingFace, headlined by EstLLM 70B, continued-pretrained from Llama 3.1 on ~60B tokens and funded by the Estonian Ministry of Education and Research under the Estonian Language Technology Programme 2018–2027 (EKTB104), jointly with TalTechNLP. Note the licence constraint — because the flagship is a Llama derivative it carries the Llama 3.1 Community License, not a permissive one; the newer Apertus-based variant is the hedge against that. No commercial API. Estonia's money in 202…
Record sources
- https://huggingface.co/tartuNLP
- https://huggingface.co/tartuNLP/Llama-3.1-EstLLM-70B-Instruct-0826
- https://arxiv.org/html/2603.02041v1
- https://arxiv.org/html/2510.21193v2
- https://baltic-review.com/estonia-eestiai-ai-training-procurement-2026/
- https://regulations.ai/regulations/RAI-EE-NA-ADAPKXX-2024
- https://github.com/TartuNLP