Claude Opus 5 Is Here: Frontier Performance at Half the Price of Fable 5
Anthropic shipped Claude Opus 5 on July 24 with pricing unchanged from Opus 4.8: $5 per million input tokens, $25 per million output. The headline feature is not a benchmark. It is a dial that lets you decide how hard the model thinks.
Anthropic released Claude Opus 5 today, calling it a model that “comes close to the frontier intelligence of Claude Fable 5 at half the price.” It is the company’s fourth model release in under two months, following Mythos 5, Fable 5 and Sonnet 5 in June.
The price did not move
Opus 5 costs exactly what Opus 4.8 cost: $5 per million input tokens and $25 per million output tokens. A faster serving mode runs at roughly 2.5 times default speed for double the price. In the API the model answers to claude-opus-5 and ships across all platforms with no data retention requirements for general access.
On subscriptions, Opus 5 becomes the default model on Claude Max and the strongest model available on Claude Pro. If you were already paying for either, you got an upgrade this morning without touching your billing settings.
Effort control is the real story
The feature worth your attention is not a leaderboard position. It is effort control: a setting that lets you dial how much work the model puts into a task, from low to maximum. That turns a decision which used to happen invisibly inside the model into an explicit, per-request trade between cost, speed and depth.
Anthropic also claims the model needs less supervision: it verifies its own work and recovers from errors without a human stepping in. If that holds up in production, it attacks the most expensive part of running agents, which has never been the tokens. It has always been the person checking every step.
What Anthropic says it does worse
Buried under the launch numbers is an admission you will not find in most coverage. Anthropic states that Opus 5 remains behind Mythos 5 on offensive cybersecurity tasks and biology research. On OSS-Fuzz both models identify vulnerabilities at a similar rate, but Opus 5 is “considerably less successful at developing exploits.” The company describes it as the most aligned model in the Opus line.
That is a deliberate product decision, not an oversight, and it tells you something useful: capability and safety posture are now tuned per model, not per generation.
What to do about it
If you run on Opus 4.8, you now get a stronger model at identical cost plus a dial to cut your bill. If you route everything through Fable 5, this is the week to measure how many of those calls actually need it. Anthropic itself recommends Fable 5 only for the most demanding work.
