Anthropic announced the model on September 28, calling it the most capable Sonnet model to date. The announcement does not specify regional availability for each platform.
Pricing and reported efficiency
The API rates match those for Sonnet 5. Anthropic says users in its internal coding and computer-use evaluations completed tasks with 30% fewer tokens on average, and that the model generated output tokens more than 30% faster. These are company-reported results; the announcement does not provide independent measurements across production workloads.
That distinction matters for budgeting: the token price is unchanged, while any per-task savings depend on how a workload performs with the model.
Benchmark results include a disclosed caveat
Anthropic reports a 70.6% score for Sonnet 5.5 on Terminal-Bench 4.0, compared with 10.3% for Sonnet 5 and 66.4% for Opus 5.5. The company says a pre-release bug affected structured-output benchmark runs and was fixed before launch. These figures are Anthropic’s reported evaluations, not independent results, and do not establish performance across other developer tasks.