Claude Opus 4.5 - Testing On Real Bugs

Claude just released Opus 4.5 - their new flagship model that's supposedly smarter AND uses 48% fewer tokens than Sonnet 4.5. I threw it at my broken SaaS kit built on Cloudflare Actors to see if it could actually fix real bugs. The app was completely broken - 31 TypeScript errors, demos not working, sign-in failing. Opus tackled it all and got everything running. Quick look at what Opus 4.5 brings: - New default recommended model in Claude Code - Higher scores on software engineering benchmarks - 48% fewer tokens at high effort vs Sonnet 4.5 - Effort parameter in the API for cost/performance control The SaaS kit I'm building uses Cloudflare's actors pattern with durable objects - still early development but interesting for exploring edge computing patterns. Links: Opus 4.5 Announcement: https://www.anthropic.com/news/claude-opus-4-5 Cloudflare Actors: https://github.com/cloudflare/actors