Anthropic Releases a Cheaper Claude, and an Outside Index Ranks It First of 212 Models
Anthropic released Claude Opus 5.5 on Sept. 22, 2026, and priced it below the model it replaces, at $4 per million input tokens and $20 per million output tokens, 20% less than Claude Opus 5.
The cut is deeper on cached text. Anthropic put cache reads, the rate charged for text already sent to the model once, at $0.20 per million tokens against $0.50 for Opus 5, and it describes those reads as most of what coding and step-by-step agent work costs. The company also said its own tests show the new model costing 40% less than Opus 5 on typical workloads at default settings. That last figure is Anthropic's own measurement of its own product.
The outside measurement comes from Artificial Analysis, an evaluation firm that runs its own tests rather than reporting figures supplied by model makers. It scores the model 58 on the Artificial Analysis Intelligence Index v4.3.2, a composite of ten evaluations, and places it first of the 212 models in its comparison class. The score belongs to one configuration, the adaptive-reasoning version run at maximum effort. Artificial Analysis also records a context window of one million tokens.
Anthropic published a table of results from benchmarks it ran itself, and a caveat alongside it: the company wrote that at these capability levels, benchmark margins have become a less reliable guide to real differences between models. It said the model was tested before release by outside evaluators including Frontier Design and METR.
The model runs on Amazon Web Services, Google Cloud and Microsoft Azure as well as Anthropic's own platform, where developers address it as claude-opus-5-5. Anthropic said Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks.
