Claude Sonnet 5.5 Gets Faster At The Same Token Price

Anthropic has released Claude Sonnet 5.5, positioning it as a quicker and more efficient model for the work people actually hand to an AI assistant. The company says the new version is more than 30% faster than Sonnet 5 and can cost up to 30% less per completed task. That last figure does not mean Anthropic has cut its API token prices; it means the model may need fewer tokens and less back-and-forth to finish some jobs.

The distinction matters for developers and businesses budgeting for AI. Anthropic lists Sonnet 5.5 at $2 per million input tokens and $10 per million output tokens, the same published rates as Sonnet 5. A shorter route to a finished answer could reduce a customer’s bill, but the saving will depend on the task, prompts and how the model is used. It should not be read as a blanket 30% discount.

According to Anthropic’s launch announcement, Sonnet 5.5 improves at coding, working through documents and producing spreadsheets and presentations. The company says it can approach the performance of its higher-end Opus 5.5 on more bounded tasks. Those are Anthropic’s own evaluations, so independent testing will be important before teams assume the same results in their workflows.

The practical pitch is easy to understand. Companies increasingly want a model that can draft, revise and check a piece of work without requiring the most expensive option for every step. Faster responses also matter when an assistant is connected to tools and must make several decisions before it can finish a request. For software teams, that could mean less waiting during coding and review; for office users, it could mean quicker document and data work.

Sonnet 5.5 is available in Claude’s apps and through the API as claude-sonnet-5-5, with access through Amazon Web Services, Google Cloud and Microsoft as well. Anthropic says it has added safeguards for risky cyber requests, including routing some high-risk tasks to Sonnet 5. The company has been expanding Claude’s role beyond chat, and its use in model development shows how much depends on dependable performance, not just benchmark headlines.

For users, this is a useful upgrade rather than a new product category. The key question is whether Sonnet 5.5’s claimed speed and efficiency hold up in day-to-day projects. If they do, the biggest benefit may be fewer stalled workflows and a lower cost for finished work, even though the token price on the rate card has not moved.