Benchmarks · creative
Judgemark v4
Meta-evaluation that rates LLM judges on how discriminatively they score fixed creative-writing samples. Maintained by EQ-Bench. Score: Composite judge-separation score on the reported 100-point scale (unclamped bounds not verified).
40 models · 40 results · data as of 2026-10-29 · eqbench.com ↗
Leaderboard
Opus 4.6 max
All results
40 rows| # | Model | Effort | Judgemark v4 |
|---|---|---|---|
| 1 | Opus 4.6 | max | 90.7 |
| 2 | GPT-5.5 | xhigh | 87.8 |
| 3 | Opus 4.7 | max | 84.0 |
| 4 | Gemini 3.7 Flash | max | 83.4 |
| 5 | Sonnet 4.6 | max | 82.1 |
| 6 | Muse Spark 1.2 | max | 79.4 |
| 7 | Opus 5 | max | 78.8 |
| 8 | Gemini 3.1 Pro | high | 78.7 |
| 9 | Opus 4.8 | max | 78.0 |
| 10 | Grok 4.5 | max | 77.1 |
| 11 | GLM 5.2 | max | 73.2 |
| 12 | GLM 5.3 | max | 72.6 |
| 13 | Gemma 4 31B IT | high | 72.3 |
| 14 | GPT-5.4 | xhigh | 72.1 |
| 15 | Sonnet 5 | max | 70.9 |
| 16 | Gemini 3.5 Flash | xhigh | 68.8 |
| 17 | Qwen 3.8 27B | max | 67.4 |
| 18 | GLM 5.1 | max | 67.2 |
| 19 | Qwen 3.5 27B | high | 60.5 |
| 20 | Gemini 3.1 Flash-Lite | high | 58.8 |
| 21 | MiMo V2.5 Pro | xhigh | 57.9 |
| 22 | Kimi K2.6 | max | 57.3 |
| 23 | Qwen 3.5 397B-A17B | max | 54.2 |
| 24 | Qwen 3.6 27B | max | 53.4 |
| 25 | Gemma 4 26B | xhigh | 53.0 |
| 26 | Nemotron 3 Ultra | max | 52.9 |
| 27 | Grok 4.3 | xhigh | 49.6 |
| 28 | DeepSeek V4 Pro | max | 47.1 |
| 29 | Gemini 3 Flash | high | 46.1 |
| 30 | Qwen 3.6 Max | max | 45.1 |