Tihu · Pelican Bench

Models: 10 · Tiers: 26 · Samples: 80 · Invalid: 2

Day highlights

GPT-6.1 Sol

8.57 · max (best)
Model page

max

8.57 ± 0.24

medium

8.18 ± 0.43
Judge scores vary between passes

low

8.10 ± 0.39

GPT-6 Astra

8.53 · max (best)
Model page

max

8.53 ± 0.30

medium

8.37 ± 0.48

low

7.80 ± 0.32

GPT-6 Sol

7.52 · max (best)
Model page

max

7.52 ± 0.56

medium

7.25 ± 0.56
Judge scores vary between passes

low

6.05 ± 1.03

Grok 4.6

7.49 · high (best)
Model page

high

7.49 ± 0.57

GPT-5.6 Sol

7.26 · max (best)
Model page

max

7.26 ± 0.27

medium

6.58 ± 0.73

low

5.48 ± 0.12

GPT-6 Luna

7.15 · max (best)
Model page

max

7.15 ± 0.46

medium

7.03 ± 0.38

low

6.25 ± 0.54

Grok 4.7

6.66 · xhigh (best)
Model page

xhigh

6.66 ± 1.61

low

6.40 ± 1.04

medium

5.41 ± 0.86

GPT-5.6 Terra

6.53 · max (best)
Model page

max

6.53 ± 0.82

low

5.99 ± 1.50

medium

5.28 ± 1.87

GPT-5.6 Luna

6.12 · max (best)
Model page

max

6.12 ± 1.61
Judge scores vary between passes

low

4.38 ± 0.44

medium

3.98 ± 0.22

Grok 4.5

5.67 · high (best)
Model page

high

5.67 ± 0.77