Claude Opus 5 Ships: Near-Fable Coding at Half the Price
Opus 5 is the new default model in Claude Code at the same $5/$25 pricing as Opus 4.8, scoring within 0.5% of Fable 5 on CursorBench 3.2 for half the cost per task.
If you use Claude Code for diffs, reviews, or refactors, you just got a free upgrade. Opus 5 is now the default on Claude Max and rolling into Claude Code at the same $5/$25 pricing as Opus 4.8. On CursorBench 3.2 at max effort, it scores within 0.5% of Fable 5's peak at half the cost per task. You can stop saving Fable for the hard stuff; Opus 5 handles it.
The 1M token context window lets you throw an entire monorepo at it without chunking. The 128k output limit handles multiple files in one shot, so you can ask Claude Code to scaffold a whole feature instead of one component at a time. Crank up effort settings when you need Fable-grade reasoning on a gnarly refactor, or dial them down for routine tasks and save tokens.
Two beta features ship alongside the model. Mid-conversation tool changes preserve your prompt cache, so you can swap tools mid-session without losing context. Automatic fallbacks route safety-flagged requests to another model instead of hard-blocking, which keeps your CI pipeline from stalling on a false positive. Update your config to claude-opus-5 today. Same cost, better model. No reason to wait.