Models · The Decoder ·
Claude Fable 5 outpaces GPT-5.5 by 13 points on FrontierMath's toughest problems
Anthropic's Claude Fable 5 reportedly scores 88% on FrontierMath's hardest tier, compared with about 75% for OpenAI's GPT-5.5. The article says this is a sharp improvement from Opus 4.5, which was below 10% earlier in 2026.