Sonnet 5 falls short on prototypes but dominates PRDs in 64‑run benchmark
Lenny’s How I AI Bench pitted Anthropic’s Claude Sonnet 5 against four rivals across PRD drafting, prototype generation, and agent tasks. The 64‑generation blind test flipped Anthropic’s hype: Sonnet 5 lags in complex prototypes but shines in PRDs and daily agent chats. The results give concrete model‑by‑task recommendations for product teams.