Claude Opus 5.5 Lands at 40 Percent Less Cost Than Opus 5

Anthropic released Claude Opus 5.5 on 22 September, its first new model since the company publicly called for pacing the frontier rather than racing it. The timing matters as much as the model.

Opus 5.5 performs at roughly the level of Claude Fable 5.1 on most work, according to Anthropic, while costing 40 percent less to run than Opus 5 on typical workloads. API pricing drops 20 percent on input and output tokens and 60 percent on cache reads. For anyone building on the API rather than just chatting with it, that pricing change is the story: a model this capable at this price shifts what becomes affordable to run at scale, from long agent sessions to document heavy research tasks.

Anthropic also says Opus 5.5 is the strongest performing model the company has tested on its automated behavioural audit, with external evaluators including METR reviewing it before release. Claude Sonnet 5.5 and Claude Haiku 5.5 are due in the coming weeks, carrying the same performance, efficiency and safety work down through the smaller models. Those two releases will matter more for most teams day to day, since Opus tends to sit behind the hardest problems rather than routine work, but Anthropic has not given a firm date beyond weeks.

We use Claude daily inside our own production pipeline, from first drafts of client copy to code review on our web builds, so a meaningful drop in cost per task is not an abstract benefit. It is the difference between running a long research or QA pass on every project and rationing it for the ones that can justify the spend.

The wider signal is worth watching too. A company that spent much of 2026 talking about slowing down just shipped a model that is both more capable and cheaper than its predecessor. Anthropic's argument is that pacing the frontier does not mean pausing it, and Opus 5.5 is the model built to prove that point.