Grouped per chain, all nineteen intervals overlap: the honest read is a tie at the top, not a ranking
Hover for intervals · click for details
all 19 intervals overlap
25%
50%
75%
100%
gpt-5.5 · 72.7% · 90% world-cluster CI [54.5, 90.9] · leads the point estimate
gpt-5.5
72.7
opus-4.8 · 69.1% · 90% world-cluster CI [47.3, 90.9]
opus-4.8
69.1
deepseek-v4-flash · 69.1% · 90% world-cluster CI [47.3, 90.9]
deepseek-v4-flash
69.1
glm-5p2 · 65.5% · 90% world-cluster CI [41.8, 90.9]
glm-5p2
65.5
qwen3p7-plus · 65.5% · 90% world-cluster CI [40.0, 90.9]
qwen3p7-plus
65.5
deepseek-v4-pro · 63.6% · 90% world-cluster CI [36.4, 90.9]
deepseek-v4-pro
63.6
qwen3p6-plus · 63.6% · 90% world-cluster CI [41.8, 87.3]
qwen3p6-plus
63.6
gpt-5.6-terra · 63.6% · 90% world-cluster CI [36.4, 90.9]
gpt-5.6-terra
63.6
haiku-4.5 · 63.6% · 90% world-cluster CI [36.4, 90.9]
haiku-4.5
63.6
sonnet-4.6 · 61.8% · 90% world-cluster CI [36.4, 81.8]
sonnet-4.6
61.8
kimi-k2p7-code · 61.8% · 90% world-cluster CI [36.4, 87.3]
kimi-k2p7-code
61.8
sonnet-5 · 60.0% · 90% world-cluster CI [38.2, 81.8]
sonnet-5
60.0
kimi-k2p6 · 58.2% · 90% world-cluster CI [34.5, 81.8]
kimi-k2p6
58.2
minimax-m2p7 · 58.2% · 90% world-cluster CI [34.5, 80.0]
minimax-m2p7
58.2
gpt-oss-20b · 56.4% · 90% world-cluster CI [30.9, 81.8]
gpt-oss-20b
56.4
kimi-k2p5 · 56.4% · 90% world-cluster CI [30.9, 81.8]
kimi-k2p5
56.4
gpt-oss-120b · 52.7% · 90% world-cluster CI [27.3, 78.2]
gpt-oss-120b
52.7
gpt-5-mini · 50.9% · 90% world-cluster CI [30.9, 72.7]
gpt-5-mini
50.9
gpt-5.4-mini · 43.6% · 90% world-cluster CI [23.6, 63.6]
gpt-5.4-mini
43.6
Point estimate (dot) · 90% world-cluster interval (whisker) · shaded band = where all 19 intervals overlap.