OptionProbability
Inkling (or any Thinking Machines model)
Qwen 4 (or any Qwen model)
Gemini 4 Carbon
GPT-6.2 Astra (6.x counts)
Claude Fable 5.5 (5.2+ counts)
GLM-5.4 (or any GLM model)
GPT-6.2 Sol
Muse Wonder 1.x
Nemotron 4 (or any Nemotron model)
Any Nano Banana model
GPT-6.2 Luna (6.x counts)
MiMo V2.7 / V3 (or any MiMo model)
Grok 4.8 (or any Grok/Composer model)
Claude Opus 5.6
DeepSeek V4.1 Pro / V5 (or any DeepSeek model)
MiniMax M3.2 (or any MiniMax model)
Any Mistral model
Gemini 4 Flash-Lite
Muse Spark 1.x/2.x
Claude Haiku 5.6
Gemini 4 Flash
Kimi K3.2 / K4 (or any Kimi model)
GPT-6.2 Terra
Claude Sonnet 5.6
Beam 1.x (or any Reflection model)
Any Safe Superintelligence (SSI) model
100
96
93
89
81
80
78
71
68
65
63
62
57
53
48
40
31
25
24
20
18
17
6
OptionVotes
YES
NO
3477
288
3092
324
422
165
Neither
Pixel loss
Embedding Loss
38
35
27
