OpenCompass LLM - Language (CompassBench 2504): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 235B A22B (Thinking) | 81 |
| 2 | Gemini 2.5 Pro (Preview 05-06) | 80.6 |
| 3 | ERNIE-X1-Turbo-32K | 80.4 |
| 4 | Doubao-1.5-thinking-pro-250415 | 79.6 |
| 5 | DeepSeek R1 | 78.9 |
| 6 | Qwen 3 32B (Thinking) | 77.7 |
| 7 | Grok 3 | 75.8 |
| 8 | Qwen 3 235B A22B | 75.4 |
| 9 | DeepSeek V3 (0324) | 74.1 |
| 10 | Gemma 3 27B (IT) | 73 |
| 11 | O4 Mini (High) | 70.4 |
| 12 | GPT-4.1 (2025-04-14) | 66.8 |
| 13 | O3 Mini (High) | 65.2 |
| 14 | GLM-4 32B (0414) | 64.7 |
| 15 | Step 2 16K | 64.4 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-llm-language-compassbench-2504 · How It Works · Data refreshed daily, snapshot 2026-09-19.