OpenCompass Reasoning - Puzzle - Chinese (CompassBench 2507): leaderboard
OpenCompass CompassBench 2507 edition (file dated 2025-08-12): logic puzzles sub-task of the reasoning category, Chinese questions. CompassBench is OpenCompass's periodic evaluation of chat models on a private question set; each edition is a separate board. Edition 2510 republished 6 of these results for the same task set; those models appear on that board only. Score 0-100, higher is better.
Metric: Score (%). Source: rank.opencompass.org.cn. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 235B A22B 2507 (Thinking) | 92 | #263 |
| 2 | Qwen 3 235B A22B 2507 Instruct | 89.7 | #315 |
| 3 | GLM-4.5 | 88.5 | #283 |
| 4 | O4 Mini (2025-04-16) (High) | 88.5 | #183 (O4 Mini (2025-04-16)) |
| 5 | Doubao-Seed-1.6-thinking-250715 | 88.5 | #169 |
| 6 | Claude Sonnet 4 (Thinking) | 87.4 | #207 (Claude Sonnet 4) |
| 7 | Kimi K2 | 86.2 | #252 |
| 8 | DeepSeek V3 (0324) | 86.2 | #349 |
| 9 | O3 (2025-04-16) (High) | 86.2 | #125 (O3 (2025-04-16)) |
| 10 | GLM 4.5 Air | 85.1 | #403 |
| 11 | Hunyuan-T1-20250711 | 83.9 | #277 |
| 12 | GPT-5 | 82.8 | #105 |
| 13 | ERNIE-X1-Turbo-32K | 81.6 | #288 |
| 14 | GPT-4.1 (2025-04-14) | 80.5 | #257 |
| 15 | MiniMax M1 80k | 70.1 | #437 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-reasoning-puzzle-chinese-compassbench-2507 · How It Works · Data refreshed daily, snapshot 2026-11-05.