JudgeBench Knowledge: leaderboard
JudgeBench knowledge slice evaluates judge-model accuracy on MMLU-Pro-derived knowledge comparisons.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 52 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 91.88 | #58 |
| 2 | GPT-5.5 | 88.96 | #31 |
| 3 | Qwen 3.6 Plus | 88.96 | #120 |
| 4 | Kimi K2.5 | 88.31 | #144 |
| 5 | Gemma 4 31B (IT) | 88.31 | #176 |
| 6 | GLM-5.1 FP8 | 88.31 | #109 |
| 7 | Qwen 3.5 397B A17B | 87.99 | #146 |
| 8 | Qwen 3.5 35B A3B | 87.99 | #256 |
| 9 | DeepSeek V3.2 Speciale | 87.99 | #128 |
| 10 | Kimi K2.6 | 87.66 | #104 |
| 11 | Qwen 3.5 122B A10B | 86.36 | #173 |
| 12 | Qwen 3.6 35B A3B | 84.74 | #226 |
| 13 | GLM-5 FP8 | 84.74 | #143 |
| 14 | Kimi K2 (Thinking) | 83.12 | #243 (Kimi K2) |
| 15 | GLM-4.7 FP8 | 82.14 | #209 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=judgebench-knowledge · How It Works · Data refreshed daily, snapshot 2026-10-20.