OpenCompass Research - MMLU-Pro: leaderboard
OpenCompass research-table score on MMLU-Pro, a harder multitask knowledge benchmark with expanded answer choices.
Metric: Score (%). Source: rank.opencompass.org.cn. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 89.3 | #66 |
| 2 | Qwen 3.5 397B A17B | 87.6 | #146 |
| 3 | Kimi K2.5 | 86.2 | #144 |
| 4 | GPT-5 (2025-08-07) (High) | 86.2 | |
| 5 | Gemini 2.5 Pro | 85.8 | #152 |
| 6 | DeepSeek V3.2 | 85.8 | #205 |
| 7 | DeepSeek V3.2 Speciale | 85.5 | #128 |
| 8 | GLM-5 | 85.2 | #141 |
| 9 | Kimi K2 (Thinking) | 84.3 | #243 (Kimi K2) |
| 10 | GLM-4.7 | 84 | #192 |
| 11 | Qwen 3 235B A22B 2507 (Thinking) | 83.5 | #260 (Qwen 3 235B A22B 2507) |
| 12 | DeepSeek R1 0528 | 83.5 | #222 |
| 13 | Step 3.5 Flash | 83.5 | #270 |
| 14 | MiMo-V2-Flash | 83.1 | #372 |
| 15 | O3 (2025-04-16) | 83 | #121 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-research-mmlu-pro · How It Works · Data refreshed daily, snapshot 2026-10-20.