Anthropic released Claude Sonnet 5.5 on 28 September. The company says it is available in its apps and API, while AWS separately confirms access through Amazon Bedrock. This is an actual release rather than a preview. Anthropic positions it between the more capable Opus 5.5 and the forthcoming lower-cost Haiku 5.5, particularly for well-scoped everyday tasks, code fixes, and producing documents or spreadsheets.

The API tariff is unchanged from Sonnet 5: $2 per million input tokens and $10 per million output tokens. Anthropic says output generation is more than 30% faster and that fewer tokens are needed for many tasks, bringing cost per completed task down by up to 30%. That is not a blanket cut in token prices. The company also reports large gains on selected coding benchmarks; such scores reflect specific test settings, not a guarantee for a real project.

If independent production tests confirm the efficiency, Sonnet 5.5 could make assistants for sorting company documents, drafting initial analyses, or helping developers with modest code changes more economical. The practical value may lie less in a record score than in reducing the wait during repeated human–AI exchanges. Anthropic itself still describes Opus 5.5 as stronger for complex, open-ended judgment.

Before deployment, teams need to measure accuracy, correction rates, end-to-end task cost, and handling of sensitive data on their own material. Safety controls do not remove the need to review outputs. Because the model is already available, limited pilots could optimistically be evaluated within a few weeks to three months; wider adoption depends on those results and has no defensible fixed date. The image is a generic editorial illustration only.