OpenCompass LLM - Reasoning (CompassBench 2507): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 66 |
| 2 | O3 (2025-04-16) (High) | 66 |
| 3 | Gemini 2.5 Pro | 65.4 |
| 4 | Doubao-Seed-1.6-thinking-250715 | 65.4 |
| 5 | Grok 4 | 63.5 |
| 6 | O4 Mini (2025-04-16) (High) | 63.5 |
| 7 | GLM-4.5 | 63.2 |
| 8 | Qwen 3 235B A22B 2507 (Thinking) | 62.8 |
| 9 | Claude Sonnet 4 (Thinking) | 60.8 |
| 10 | DeepSeek R1 0528 | 60.3 |
| 11 | Hunyuan-T1-20250711 | 59.4 |
| 12 | Qwen 3 235B A22B 2507 Instruct | 58.7 |
| 13 | Kimi K2 | 57.9 |
| 14 | GLM 4.5 Air | 57.5 |
| 15 | ERNIE-X1-Turbo-32K | 57.1 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-llm-reasoning-compassbench-2507 · How It Works · Data refreshed daily, snapshot 2026-09-19.