CroissantLLMChat-v0.1: benchmark results
CroissantLLM's 1.3B bilingual English-French chat model, pretrained on a 1:1 English-to-French token ratio with a custom tokenizer, checkpointed at 190,000 steps (2.99T tokens) plus a chat fine-tuning phase. Provider: CroissantLLM. Released 2024-01-24. Access: Open.
Aggregate Rating 1038 ± 32, rank #1638 of 1640 rated models, from 54 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CARTE - Occitanie | 25.9 | Accuracy (%) on the 139 CARTE questions about Occitanie, 0-s | 17.3 |
| EuroEval Portuguese NLU - ScaLA PT | 0.61 | Linguistic acceptability Score (%) | 16.1 |
| Kurakurai French LLM - AIME 2025 | 0 | Accuracy (%) | 15.9 |
| EuroEval Italian NLU - SQuAD IT | 20.31 | Reading comprehension Score (%) | 15.5 |
| CARTE - Pays de la Loire | 18.8 | Accuracy (%) on the 133 CARTE questions about Pays de la Loi | 13.5 |
| Kurakurai French LLM - AIME 2024 | 0 | Accuracy (%) | 13.4 |
| EuroEval Spanish NLU - MLQA ES | 17.67 | Reading comprehension Score (%) | 13 |
| EuroEval Italian Knowledge | 0.65 | Knowledge Average Score (%) | 12.9 |
| EuroEval Portuguese NLU - MultiWikiQA PT | 12.1 | Reading comprehension Score (%) | 12 |
| EuroEval Spanish NLU - ScaLA ES | 0 | Linguistic acceptability Score (%) | 11.7 |
| CARTE - Corse | 18 | Accuracy (%) on the 100 CARTE questions about Corse (Corsica | 11.5 |
| CARTE - Normandie | 21.8 | Accuracy (%) on the 156 CARTE questions about Normandie, 0-s | 11.5 |
Interactive version: theaggregate.ai/model?slug=croissantllmchat-v0-1 · How It Works · Data refreshed daily, snapshot 2026-10-13.