Anthropic releases Claude Sonnet 5.5 at the same price, and gives a mid-tier model its top-tier cyber safeguards

Anthropic released Claude Sonnet 5.5 on 28 September, calling it the second model in the Claude 5.5 family. The list price does not change: $2 per million input tokens and $10 per million output tokens, with $0.20 per million for cache reads. It is available on Anthropic's own platform and on Amazon Web Services, Google Cloud and Microsoft Azure.

The performance claims are Anthropic's own. It says the model runs more than 30% faster than Sonnet 5 and costs up to 30% less per task because it uses fewer tokens. Its comparison table puts Sonnet 5.5 at 70.6% on Terminal-Bench 4.0 against 66.4% for Opus 5.5, and at 1844 on GDPval-AA v2.1 against 1846 for Opus 5.5 and 1449 for Sonnet 5. None of that is independently measured yet, and the page notes results vary with the effort setting.

The safety line is the more unusual part. Anthropic writes that “Sonnet 5.5's cyber capabilities are a large improvement over Sonnet 5's, so we're deploying it with safeguards similar to those on Opus 5.5.” In practice, higher-risk cybersecurity tasks will visibly fall back to Sonnet 5, while routine bug-fixing is unaffected. Anthropic says it is the first Sonnet to launch with cyber safeguards of this kind.

The llm-stats release tracker lists it as the only model release dated inside this window. A system card is linked from the launch page; nothing from it is reported here because it could not be opened in this run.

• Confirmed Released 28 September at $2 per million input tokens and $10 per million output tokens, on Anthropic's platform, AWS, Google Cloud and Azure. Anthropic
• Claimed More than 30% faster than Sonnet 5 and up to 30% cheaper per task; 70.6% on Terminal-Bench 4.0 against 66.4% for Opus 5.5, on Anthropic's own evaluations. Anthropic
• Confirmed Deployed with cyber safeguards similar to Opus 5.5; higher-risk cybersecurity tasks visibly fall back to Sonnet 5. Anthropic