Models: 23 · Tiers: 48 · Samples: 148 · Invalid: 4

Day highlights

Qwen3.8-Max

8.89 · default (best)
Model page

Gemini 3.8 Flash

8.82 · high (best)
Model page

medium

8.46 ± 0.39
Judge scores vary between passes

low

8.02 ± 0.78
Judge scores vary between passes

GPT-6.1 Sol

8.79 · max (best)
Model page

GPT-6 Astra

8.67 · max (best)
Model page

Claude Opus 5.5

8.23 · medium (best)
Model page

Kimi K3

8.22 · max (best)
Model page

GPT-6 Sol

8.15 · max (best)
Model page

GLM 5.3

8.11 · default (best)
Model page

GPT-6 Luna

7.82 · max (best)
Model page

low

6.23 ± 0.99
Judge scores vary between passes

GPT-5.6 Sol

7.28 · max (best)
Model page

Gemini 3.1 Pro

6.91 · default (best)
Model page

GPT-5.6 Terra

6.65 · max (best)
Model page

low

5.23 ± 1.18

DeepSeek V4 Pro

6.62 · default (best)
Model page

default

6.62 ± 0.09
Judge scores vary between passes

Claude Sonnet 5.5

6.54 · high (best)
Model page

Grok 4.6

6.46 · high (best)
Model page

GPT-5.6 Luna

6.45 · max (best)
Model page

Doubao Seed 2.1 Pro

6.41 · default (best)
Model page

MiniMax-M3

6.37 · default (best)
Model page

Hy3

6.09 · default (best)
Model page

GLM 5.2

5.99 · max (best)
Model page

Grok 4.7

5.80 · medium (best)
Model page

xhigh

5.22 ± 0.79

Grok 4.5

5.70 · high (best)
Model page

Composer 2.5

4.99 · default (best)
Model page

default

4.99 ± 1.55
Judge scores vary between passes