105 real bugs: how Fable 5.2 stacks up on cost and fixes
Paweł Huryn compares Fable, Sonnet and GPT-6 models on two repos.

Paweł Huryn (@PawelHuryn) tested models on real work: two repos with 105 bugs, asking each to find and fix what it could.
His results: Fable 5.2 (max) fixed 58 for $139.65, Sonnet 5.5 (max) 51.3 for $154.03, GPT-6 Astra (max) 45 for $33.03, GPT-6.1 Sol (max) 44.3 for $5.88, and Fable 5.1 (max) 43 for $87.18 (single run).
Why it mattersIt puts numbers on the trade-off between the best score and the cheapest fix.
The post drew about 528 likes by Tuesday morning.



