“Matt Schumer's testing indicated that for almost every task he tested, Fable was "quite a bit better" and more agentic than GPT-5.6 SOL.”