LLM Stats (MMLU-Redux): leaderboard
Re-annotated subset of 3,000 MMLU questions across 30 subjects corrected for labeling errors and ambiguous options, providing more reliable broad knowledge measurement.
Metric: Score (%). Source: llm-stats.com. Source: qwen.ai. Saturation forecast: Estimated already saturated. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.6 (Thinking) | 95.3 | #104 (Kimi K2.6) |
| 2 | Claude Opus 4.6 (Thinking) | 95.2 | #63 (Claude Opus 4.6) |
| 3 | Qwen 3.7 Max | 95 | #75 |
| 4 | Qwen 3.5 397B A17B | 94.9 | #146 |
| 5 | DeepSeek V4 Pro (Reasoning, Max Effort) | 94.8 | #102 (DeepSeek V4 Pro) |
| 6 | Qwen 3.6 Plus | 94.5 | #120 |
| 7 | Qwen 3.7 Plus | 94.5 | #92 |
| 8 | GLM-5.1 (Thinking) | 94.3 | #97 (GLM-5.1) |
| 9 | Qwen 3.5 122B A10B | 94 | #173 |
| 10 | Qwen 3 235B A22B 2507 (Thinking) | 93.8 | #260 (Qwen 3 235B A22B 2507) |
| 11 | Qwen 3 VL 235B A22B (Thinking) | 93.7 | #238 (Qwen 3 VL 235B A22B) |
| 12 | Qwen 3.6 27B | 93.5 | #170 |
| 13 | DeepSeek R1 0528 | 93.4 | #222 |
| 14 | Qwen 3.5 35B A3B | 93.3 | #256 |
| 15 | Qwen 3.6 35B A3B | 93.3 | #226 |
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-redux · How It Works · Data refreshed daily, snapshot 2026-10-20.