Claude Sonnet 5.5 costs the same. So why is it cheaper?

Anthropic has released Claude Sonnet 5.5 across Claude, its API, AWS, Google Cloud, and Microsoft Azure. GitHub has also made it generally available in Copilot, rolling out gradually across its IDEs, CLI, coding agent, app, web, and mobile clients.

Anthropic reports that Sonnet 5.5 generates output more than 30% faster than Sonnet 5 and usually costs up to 30% less per completed task. The list prices stay the same: $2 per million input tokens and $10 per million output tokens.

The key difference is task cost, not token price. A workflow can cost less to finish if it uses fewer tokens, needs fewer retries, or completes sooner.

The API model ID is claude-sonnet-5-5.

There is one migration detail. Integrations that disable thinking must adopt the new between_tools setting. Test behavior between tool calls rather than treating this as a simple model-ID swap.

For routine coding and document work, run a small A/B test against your current model on the same tasks, and record:

1. Completion time
2. Retry count
3. Actual token cost
4. Output quality

The catch: the speed and efficiency numbers are Anthropic’s own benchmarks and have not been independently checked. Pick your default from your own results, not the launch numbers alone.