Anthropic announces Claude Opus 5.5 at lower cost

Anthropic announced Claude Opus 5.5 on 2026-09-22, describing it as the first model in a new Claude 5.5 family. The company says it performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5. The announcement comes from anthropic.com.

What Anthropic says changed

Two claims carry the announcement. The first is capability: Anthropic describes Opus 5.5 as a major step up from Opus 5 and calls it the new leading model, adding that early testers saw large jumps in performance on their most complex work. The second is price. Anthropic says its tests show that at default settings the model costs 40% less than Opus 5 on typical workloads, with input tokens at $4 and output tokens at $20 per million, 20% less than Opus 5.

Those two claims point in different directions, and that is the interesting part. A cheaper model at the same tier changes the arithmetic of work that was previously rationed. A stronger model at the same tier changes which tasks are worth attempting at all. Anthropic is claiming both at once, which is why the announcement is worth reading rather than skimming.

Safety tier and what follows

Anthropic says Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, and that it is deploying the model with safeguards similar to those on Claude Fable 5.1. The claimed capabilities and the controls around their use both matter when evaluating the model.

Anthropic also says Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks, with many of the same improvements to performance, efficiency and safety. No dates are given for those two.

For teams already running work on Opus 5, the practical question is narrow: does the same task produce the same or better output at the stated lower token price? That is testable on your own workloads, and it is the only test that matters for a budget. The 40% figure is Anthropic's own measurement on what it calls typical workloads; your mix of input and output tokens will differ, and the $4 and $20 per million rates are the numbers you can actually calculate against.

Anthropic’s announcement includes benchmark results and named customer testimonials. Those provide context for its claims, but they do not guarantee the same result on every task. A small comparison using your own work can show whether the gains transfer: assess output quality and cost per completed task before changing a production system.

The follow-on models matter for planning. If Sonnet 5.5 and Haiku 5.5 arrive in the coming weeks with similar improvements, a decision made now on Opus 5.5 pricing may look different within a month. Worth knowing before committing a large workload to one tier.