Qwen 2.5 3B Instruct: benchmark results
Alibaba Qwen 2.5 3B instruction-tuned checkpoint. Provider: Alibaba. Released 2024-09-19. Access: Open.
Aggregate Rating 1416 ± 7, rank #1162 of 1656 rated models, from 667 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LA Leaderboard - WNLI Spanish | 70.42 | Accuracy (%) | 94.1 |
| Open CoT - LogiQA | 9.42 | CoT Gain (%) | 93.5 |
| SiT-Bench - Global Perception & Mapping | 27.93 | Accuracy (%; item-weighted over the category's subtasks) | 93.3 |
| LA Leaderboard - WNLI Catalan | 66.2 | Accuracy (%) | 91.2 |
| LA Leaderboard - NoticIA | 28.36 | ROUGE-1 (%) | 90.8 |
| LA Leaderboard - GalCoLA | 16.12 | MCC (x100) | 89.7 |
| LA Leaderboard - Belebele Catalan | 80.89 | Accuracy (%) | 88.2 |
| LA Leaderboard - Belebele Spanish | 84 | Accuracy (%) | 88.2 |
| LA Leaderboard - Fake News Spanish | 79.37 | Accuracy (%) | 88.2 |
| DeEscalWild - METEOR | 15.7 | METEOR (0-100) of the generated civilian turns against the r | 87.5 |
| RenoBench - Surname | 93 | Recall (%) of gold JATS surname annotations reproduced by th | 87.5 |
| RenoBench - Title | 93.9 | Recall (%) of gold JATS article and chapter title annotation | 87.5 |
Interactive version: theaggregate.ai/model?slug=qwen-2-5-3b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.