Zvi on Claude Fable 5.2: Your Daily Driver Is Fine, Hard Tasks Demand the Bazooka

TheZvi · x · 2026-11-17

Zvi's review of Claude Fable 5.2: indistinguishable from Opus 5.5 for daily use, but the go-to model for difficult, complex, or long tasks. Ethan Mollick calls it 'almost AGI' / 'jagged AGI'. Benchmarks show Fable 5.2 topping FrontierCode 2.0 at 49.1% (xHigh) vs Opus 5.5's 45.5% max and GPT-6 Astra's 42.8%, and reaching 82.1% on WANDR at Max reasoning effort.

Original post →

More from Models

Models channel →