The Key Facts
Opus 5 is available immediately across Anthropic's platforms, exposed via API as `claude-opus-5`, at $5/M input and $25/M output—the same price as the Opus 4.8 it replaces. It becomes the default model on Claude Max and the strongest tier on Claude Pro; Fast mode runs at roughly 2.5× the default speed for twice the base price. The headline pitch is "close to Fable 5 at half the price": at maximum effort on CursorBench 3.2 it scores within 0.5% of Fable 5's peak while cutting per-task cost in half, and on Frontier-Bench v0.1 it surpasses every model while more than doubling Opus 4.8's score. Every one of Anthropic's life-sciences evaluations improved over Opus 4.8, with organic chemistry up 10.2 points. On safety, it self-reports a 2.3 on overall misaligned behavior, the lowest of its recent models, though it still trails Mythos 5 on cyber-offense tasks.
One New Button
The most-watched addition is an adjustable effort dial: users pick low, medium, or high to decide how much reasoning the model spends per answer, tuning spend against quality themselves. Two API betas ship alongside it: mid-conversation tool changes that no longer invalidate the prompt cache, and automatic fallbacks that reroute safety-flagged requests to another model instead of failing outright.
Why It Matters
For developers this is a "price unchanged, capability up, now with a dial" upgrade—worth switching the default call and running a regression pass, while the effort dial exposes cost control directly to the business side for the first time. It is Anthropic's fourth model in under two months (after Mythos 5, Fable 5, and Sonnet 5), a cadence that lands as the company prepares to IPO later this year.
via: Anthropic official announcement "Introducing Claude Opus 5" (2026-07-24), with same-day reporting by CNBC, Fortune, and Bloomberg; verified 2026-07-25