OpenCompass Reasoning - Academic - Chinese (CompassBench 2510): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 (2025-04-16) (High) | 46.7 |
| 2 | GPT-5 (Thinking) | 43.3 |
| 3 | GLM-4.6 | 42.5 |
| 4 | Gemini 2.5 Pro | 41.7 |
| 5 | Claude Sonnet 4.5 (Thinking) | 38.3 |
| 6 | Grok 4 | 35.8 |
| 7 | O4 Mini (2025-04-16) (High) | 31.7 |
| 8 | Qwen 3 235B A22B 2507 (Thinking) | 29.2 |
| 9 | GPT-OSS-120B (High) | 29.2 |
| 10 | DeepSeek V3.2 Exp | 26.7 |
| 11 | Seed-OSS-36B-Instruct | 26.7 |
| 12 | DeepSeek R1 0528 | 25 |
| 13 | Hunyuan-T1-20250822 | 25 |
| 14 | Qwen 3 Next 80B A3B (Thinking) | 21.7 |
| 15 | Ling-1T | 21.7 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-reasoning-academic-chinese-compassbench-2510 · How It Works · Data refreshed daily, snapshot 2026-09-19.