I Left Claude This Summer. Opus 5.5 Is Winning Me Back.

Why watch

Developer and content creator Matt Maher compares Claude Opus 5.5 with GPT-6 Sol, Fable, and Opus 5 using a one-shot build test and his custom CARE intent benchmark. He evaluates design fidelity, token usage, cost, and communication quality for long-running agentic work to determine if Anthropic's claim of Fable-level performance at 40% lower cost holds true. The video is for developers choosing between AI coding assistants who want data-driven model selection insights.