BREAKING
4 Frontier Models, 1 Game Prompt
Intelligence Index scores
Fable 560
GLM 5.2 max51
Deepseek V4 Pro44
SWE-bench Pro coding
Fable 580
GPT-5.6 Sol64.6
GLM 5.262.1
Per-run cost splits the field
Fable 5$0.42-$0.80
Quality lead on complex tasks
Hardest to justify on price
Cheaper models$0.15 and under
GPT-5.6 around $0.15
GLM 5.2 $0.05-$0.07, Deepseek under a cent
No single model wins outright
Best experience per dollar
AI NEWS BLITZ
Four leading AI models got the same prompt to build one Contra-style game.