Tihu · Pelican Bench

Models: 10 · Tiers: 26 · Samples: 78 · Invalid: 0

Day highlights

GPT-6 Astra

8.48 · max (best)
Model page

max

8.48 ± 0.33

medium

7.67 ± 0.83
Judge scores vary between passes

low

6.99 ± 0.07
Judge scores vary between passes
Judge scores vary between passes

GPT-6.1 Sol

8.27 · max (best)
Model page

max

8.27 ± 0.45

medium

7.40 ± 0.91

low

7.31 ± 0.23

GPT-6 Luna

7.65 · medium (best)
Model page

medium

7.65 ± 0.51

max

6.51 ± 0.40
Judge scores vary between passes

low

6.10 ± 1.00

GPT-5.6 Terra

7.57 · max (best)
Model page

max

7.57 ± 0.78

low

6.80 ± 1.13
Judge scores vary between passes

medium

6.34 ± 0.34
Judge scores vary between passes

GPT-5.6 Sol

7.28 · low (best)
Model page

low

7.28 ± 0.39

max

6.94 ± 1.70
Judge scores vary between passes

medium

6.93 ± 0.92

GPT-6 Sol

7.27 · max (best)
Model page

max

7.27 ± 0.21

medium

6.72 ± 0.47

low

6.60 ± 1.76
Judge scores vary between passes

Grok 4.6

7.01 · high (best)
Model page

high

7.01 ± 0.40

Grok 4.7

6.88 · xhigh (best)
Model page

xhigh

6.88 ± 1.88

medium

5.22 ± 0.69

low

5.04 ± 0.49

Grok 4.5

6.85 · high (best)
Model page

high

6.85 ± 0.32

GPT-5.6 Luna

5.78 · low (best)
Model page

low

5.78 ± 0.56

max

5.53 ± 0.62

medium

3.73 ± 0.26