ProLLM - LLM-as-a-Judge — leaderboard
ProLLM benchmark evaluating how well LLMs can judge and score the outputs of other AI models - measuring meta-evaluation capability.
Metric: Score (%). Source: www.prollm.ai. Status: saturated. 58 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 85.3 |
| 2 | GPT-4o | 84.6 |
| 3 | GPT-OSS-120B | 84.6 |
| 4 | GPT-4 Turbo | 83.8 |
| 5 | O1 Mini | 83.8 |
| 6 | DeepSeek V3 | 82.4 |
| 7 | O1 | 82.4 |
| 8 | GPT-4.5 | 82.4 |
| 9 | O3 Mini (Medium) | 82.4 |
| 10 | O3 Mini (High) | 81.6 |
| 11 | Qwen 2.5 72B Instruct | 80.9 |
| 12 | Llama 4 Maverick Instruct | 80.9 |
| 13 | GPT-4.1 Mini | 80.1 |
| 14 | Llama 3.3 70B Instruct | 79.4 |
| 15 | Gemma 2 27B (IT) | 79.4 |
Interactive version: theaggregate.ai/benchmark?slug=prollm-llm-as-a-judge · How the rankings work · Data refreshed daily, snapshot 2026-07-22.