JudgeBench Coding: leaderboard
JudgeBench coding slice evaluates judge-model accuracy on code-answer 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) | 97.62 | #58 |
| 2 | GPT-5.5 | 97.62 | #31 |
| 3 | Kimi K2.5 | 97.62 | #144 |
| 4 | GPT-5 Mini | 97.62 | #186 |
| 5 | Gemma 4 31B (IT) | 97.62 | #176 |
| 6 | Kimi K2.6 | 97.62 | #104 |
| 7 | Qwen 3.5 35B A3B | 97.62 | #256 |
| 8 | DeepSeek R1 0528 | 97.62 | #222 |
| 9 | Step 3.5 Flash | 97.62 | #270 |
| 10 | DeepSeek V3.2 Speciale | 97.62 | #128 |
| 11 | GLM-4.7 FP8 | 97.62 | #209 |
| 12 | GLM-5 FP8 | 97.62 | #143 |
| 13 | GLM-5.1 FP8 | 97.62 | #109 |
| 14 | GPT-OSS-20B | 96.43 | #518 |
| 15 | GPT-OSS-120B | 95.24 | #341 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=judgebench-coding · How It Works · Data refreshed daily, snapshot 2026-10-20.