Anthropic releases Claude Opus 5.5, the first Claude 5.5 model, at a lower price

Anthropic says Claude Opus 5.5 matches Fable 5.1 on most tasks, and that typical workloads run 40 percent cheaper than on Opus 5.

FUFerhat UğurluYazSis TeamPublished: September 22, 20262 min read
In short

On 22 September 2026 Anthropic announced Claude Opus 5.5, the first model in its Claude 5.5 family. The company says it matches Claude Fable 5.1 on most tasks and that typical workloads cost 40 percent less to run than on Opus 5. Outside evaluators checked the model before launch.

Anthropic announced Claude Opus 5.5 on 22 September 2026. According to the company's announcement, Opus 5.5 opens the Claude 5.5 family and matches Claude Fable 5.1 on most tasks; outside groups such as Frontier Design and METR evaluated it before launch. Anthropic added that Sonnet 5.5 and Haiku 5.5 are due within the next few weeks.

  • Pricing is 4 dollars per million input tokens and 20 dollars per million output tokens, down from 5 and 25 dollars for Opus 5. According to the company's tests, cost on typical workloads at default settings is 40 percent lower.
  • In Anthropic's published results it scored 66.4 percent on Terminal-Bench 4.0, an agentic coding test, and its output is more than 30 percent faster than Opus 5.
  • The model is available in the API as claude-opus-5-5 and through Amazon Web Services, Google Cloud and Microsoft Azure, with a zero data retention option.
  • The model comes with Anthropic's watermarking measures for EU AI Act compliance, and thinking mode can no longer be turned off.

Why it matters

Lower prices for top-tier models could reduce the cost of handing heavy work such as long document analysis, code migration or report writing to AI. The zero data retention option is a point companies handling customer data may want to raise when negotiating contracts. Watermarking measures aimed at the EU AI Act could also be worth following for Turkish companies that export to Europe and deliver AI-generated output to customers there. The performance figures are Anthropic's own measurements.

Sources

FU
Ferhat Uğurlu · YazSis Team

Builds enterprise software, AI and digital transformation products together with the YazSis team.