Qwen2.5-7B-Instruct-1M: benchmark results
Alibaba's Qwen2.5-7B-Instruct-1M, a long-context variant of Qwen2.5-7B-Instruct supporting up to 1,010,000 tokens of context while maintaining short-task performance. Provider: Alibaba. Released 2025-01-27. Access: Open.
Aggregate Rating 1468 ± 22, rank #894 of 1656 rated models, from 82 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MATH Level 5 | 43.35 | Score | 92.4 |
| Open LLM Leaderboard - IFEval | 74.48 | Score | 89.6 |
| Open Japanese LLM - Jsick Exact Match | 84.29 | Score (%) | 87.7 |
| RULER | 91.8 | Avg Accuracy (%) | 81.8 |
| Open Japanese LLM - Wiki PAS SET F1 | 8.44 | Score (%) | 81 |
| Enkrypt AI - Safety Risk | 21.86 | Risk Score | 76.2 |
| Open LLM Leaderboard - BBH | 54.04 | Score | 69.1 |
| Open Japanese LLM - Jamp Exact Match | 65.23 | Score (%) | 64.8 |
| Open Japanese LLM - RC | 90.74 | Score (%) | 63 |
| Open Japanese LLM - Jnli Exact Match | 82.54 | Score (%) | 58.7 |
| Enkrypt AI - Insecure Code Risk | 24 | Risk Score | 56.1 |
| Open Japanese LLM - Jsts Spearman | 85.71 | Score (%) | 53.6 |
Interactive version: theaggregate.ai/model?slug=qwen2-5-7b-instruct-1m · How It Works · Data refreshed daily, snapshot 2026-10-20.