Models: 17 · Tiers: 42 · Samples: 129 · Invalid: 3

Day highlights

GPT-6 Astra

8.84 · max (best)
Model page

GPT-6.1 Sol

8.55 · max (best)
Model page

Kimi K3

8.24 · max (best)
Model page

Claude Opus 5.5

8.02 · high (best)
Model page

Gemini 3.8 Flash

7.88 · medium (best)
Model page

high

7.87 ± 0.37
Judge scores vary between passes

low

7.75 ± 0.79
Judge scores vary between passes

GPT-5.6 Sol

7.82 · max (best)
Model page

max

7.82 ± 0.77
Judge scores vary between passes

medium

7.22 ± 1.23
Judge scores vary between passes

GLM 5.2

7.75 · max (best)
Model page

max

7.75 ± 0.42

Claude Sonnet 5.5

7.69 · high (best)
Model page

GPT-6 Luna

7.30 · max (best)
Model page

low

6.38 ± 1.60

medium

5.30 ± 0.56
Judge scores vary between passes

Gemini 3.1 Pro

7.26 · default (best)
Model page

Grok 4.7

7.17 · xhigh (best)
Model page

GPT-6 Sol

6.95 · medium (best)
Model page

medium

6.95 ± 0.29
Judge scores vary between passes
Judge scores vary between passes

low

5.86 ± 1.17
Judge scores vary between passes
Judge scores vary between passes

GPT-5.6 Luna

6.69 · max (best)
Model page

Grok 4.6

6.62 · high (best)
Model page

Grok 4.5

6.48 · high (best)
Model page

GPT-5.6 Terra

6.25 · max (best)
Model page

Composer 2.5

4.78 · default (best)
Model page