Tihu · Pelican Bench

Models: 10 · Tiers: 26 · Samples: 81 · Invalid: 3

Day highlights

GPT-6 Astra

8.85 · max (best)
Model page

max

8.85 ± 0.23

medium

8.35 ± 0.04

low

8.35 ± 0.30

GPT-6.1 Sol

8.74 · max (best)
Model page

max

8.74 ± 0.16

medium

8.01 ± 0.51

low

7.67 ± 0.57

GPT-6 Luna

7.96 · max (best)
Model page

max

7.96 ± 0.42

medium

6.07 ± 1.58

low

5.01 ± 0.87

GPT-6 Sol

7.63 · max (best)
Model page

max

7.63 ± 0.59
Judge scores vary between passes

medium

6.81 ± 0.46

low

5.55 ± 0.33

GPT-5.6 Sol

7.15 · max (best)
Model page

max

7.15 ± 0.65

low

6.50 ± 0.64

medium

6.10 ± 0.85

GPT-5.6 Terra

7.14 · max (best)
Model page

max

7.14 ± 1.11
Judge scores vary between passes

medium

6.46 ± 1.05

low

5.12 ± 0.79

Grok 4.6

6.32 · high (best)
Model page

high

6.32 ± 1.41

Grok 4.5

6.21 · high (best)
Model page

high

6.21 ± 1.11

GPT-5.6 Luna

6.06 · max (best)
Model page

max

6.06 ± 0.67

medium

5.58 ± 0.60

low

4.53 ± 0.50

Grok 4.7

6.01 · medium (best)
Model page

medium

6.01 ± 0.59

xhigh

5.12 ± 0.82

low

4.16 ± 1.35