OpenCompass Math - Reasoning Computation - Chinese (CompassBench 2507): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 66.2 |
| 2 | Doubao-Seed-1.6-thinking-250715 | 64.4 |
| 3 | Grok 4 | 63.1 |
| 4 | GLM 4.5 Air | 60.6 |
| 5 | GLM-4.5 | 59.4 |
| 6 | Hunyuan-T1-20250711 | 59.4 |
| 7 | Claude Sonnet 4 (Thinking) | 57.5 |
| 8 | Qwen 3 235B A22B 2507 Instruct | 56.9 |
| 9 | MiniMax M1 80k | 53.8 |
| 10 | ERNIE-X1-Turbo-32K | 53.1 |
| 11 | Gemini 2.5 Pro | 50.6 |
| 12 | Kimi K2 | 48.1 |
| 13 | DeepSeek V3 (0324) | 46.9 |
| 14 | GPT-4.1 (2025-04-14) | 26.9 |
| 15 | MiniMax-Text-01 | 10 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-math-reasoning-computation-chinese-compassbench-2507 · How It Works · Data refreshed daily, snapshot 2026-09-19.