tartuNLP/Llama-3.1-EstLLM-70B-Instruct-0826

TEXT GENERATIONPricing:Input $3.5 / Cached $0.7 / Output $8.3Concurrent Unit Cost:4Model Size:70BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Aug 17, 2026License:llama3.1Architecture:Transformer Featherless Exclusive Cold

The tartuNLP/Llama-3.1-EstLLM-70B-Instruct-0826 is a 70 billion parameter instruction-following causal language model developed by TartuNLP and TalTechNLP. Continuously pre-trained from Llama-3.1-70B on 60B tokens, it is specifically optimized for high performance in Estonian language tasks, including instruction-following, language competence, and knowledge reasoning. This model also demonstrates strong capabilities in English, making it a robust bilingual solution for complex NLP applications.

Loading preview...

Model Overview

tartuNLP/Llama-3.1-EstLLM-70B-Instruct-0826 is a 70 billion parameter instruction-following model developed by the TartuNLP and TalTechNLP research groups, funded by the Estonian Ministry of Education and Research. It was continuously pre-trained from Meta's Llama-3.1-70B on approximately 60 billion tokens, followed by supervised fine-tuning and direct preference optimization. This process has significantly enhanced its capabilities in both Estonian and English.

Key Capabilities

  • Bilingual Proficiency: Excels in both Estonian and English, demonstrating strong instruction-following and reasoning abilities in both languages.
  • Estonian Language Competence: Achieves top scores in Estonian language benchmarks such as Grammar-et (0.8950), Inflection-et (0.9142), and Word-Meanings-et (0.9719).
  • Knowledge and Reasoning: Shows leading performance in Estonian knowledge and reasoning tasks, including Winogrande-et (0.8393) and GlobalPIQA-et (0.74).
  • Translation: Ranks highest in English to Estonian translation with a BLEU score of 0.278 on the wmt24pp dataset.
  • Instruction Following: Achieves a strong 0.7581 on IFEval-et and a competitive 0.9147 on IFEval-en, outperforming many comparable models.

Good For

  • Applications requiring high-quality Estonian language generation and understanding.
  • Bilingual (Estonian-English) instruction-following and conversational AI.
  • Research and development in Estonian natural language processing.
  • Tasks involving complex reasoning and knowledge retrieval in Estonian contexts.

Finetunes
1 models