LLM Stats (MMLU-ProX): leaderboard
MMLU-Pro extended to 29 languages with 11,829 questions per language, measuring advanced multi-domain reasoning across STEM, humanities, and professional fields.
Metric: Score (%). Source: llm-stats.com. Source: qwen.ai. Saturation forecast: Around January 2027. 32 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Max | 87 | #75 |
| 2 | Claude Opus 4.6 (Thinking) | 86.1 | #63 (Claude Opus 4.6) |
| 3 | Qwen 3.7 Plus | 85.4 | #92 |
| 4 | Qwen 3.5 397B A17B | 84.7 | #146 |
| 5 | Qwen 3.6 Plus | 84.7 | #120 |
| 6 | GLM-5.1 (Thinking) | 83.9 | #97 (GLM-5.1) |
| 7 | DeepSeek V4 Pro (Reasoning, Max Effort) | 83.9 | #102 (DeepSeek V4 Pro) |
| 8 | Kimi K2.6 (Thinking) | 83.7 | #104 (Kimi K2.6) |
| 9 | Nemotron 3 Ultra | 83 | #235 |
| 10 | Qwen 3.5 27B | 82.2 | #187 |
| 11 | Qwen 3.5 122B A10B | 82.2 | #173 |
| 12 | Qwen 3.5 35B A3B | 81 | #256 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 81 | #260 (Qwen 3 235B A22B 2507) |
| 14 | Qwen 3 VL 235B A22B (Thinking) | 80.6 | #238 (Qwen 3 VL 235B A22B) |
| 15 | Qwen 3 235B A22B 2507 Instruct | 79.4 | #300 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-mmlu-prox · How It Works · Data refreshed daily, snapshot 2026-10-20.