Judgemark v2.1
Chat & Writing · 2025-01-31
Earlier meta-evaluation benchmark for LLM judging ability.
Top models (higher is better)
| Model | Score |
|---|---|
| Sonnet 4.6 | 94.1 |
| Opus 4.6 | 92.9 |
| Gemini 3 Pro Preview | 86.6 |
| GLM-5 | 85.6 |
| Qwen3.5 397B A17B | 85.5 |
| Gemini 3 Flash Preview | 85.4 |
| GPT-5.4 | 85.0 |
| Sonnet 4.5 | 84.6 |
| Qwen3.5-27B | 84.3 |
| GPT-5.2 | 82.0 |
| Sonnet 4 | 82.0 |
| Qwen3.5-Flash | 81.8 |
| Qwen3.5 35B A3B | 81.4 |
| GPT-5 (2025-08-07) | 81.2 |
| GPT-5.4 Mini | 81.1 |
| Qwen3.5 122B A10B | 81.0 |
| Gemini 3.1 Flash-Lite Preview | 80.9 |
| Haiku 4.5 | 79.7 |
| GLM-4.7 | 77.5 |
| GPT-4.1 | 76.3 |
| Grok 4.1 Fast | 75.8 |
| o3 | 75.1 |
| Qwen3.5-9B | 75.0 |
| GPT-5 Mini (2025-08-07) | 74.5 |
| Kimi K2 Instruct | 73.1 |
| Gemini 2.5 Pro | 72.1 |
| Mistral Large 3 675B Instruct 2512 | 68.2 |
| DeepSeek-V3.2 | 66.3 |
| GLM 4.7 Flash | 66.2 |
| GPT-5.4 Nano | 64.5 |
| MiniMax M2.5 | 64.5 |
| GLM-4.5 | 63.4 |
| Gemini 2.5 Flash | 63.0 |
| GPT-5 Nano | 61.1 |
| Qwen3 235B A22B Instruct 2507 | 59.7 |
| Mistral Medium 3 | 58.5 |
| GPT-4.1 Mini | 57.3 |
| Qwen3-235B-A22B | 56.2 |
| O4 Mini | 56.1 |
| Gemini 2.5 Flash-Lite | 56.0 |