Anthropic says Claude Sonnet 5.5 scored 70.6% on Terminal-Bench 4.0, compared with 10.3% for Sonnet 5. The benchmark tests AI agents on multi-step command-line coding tasks.
In the company’s tests, Sonnet 5.5 generated output more than 30% faster than Sonnet 5 and cost up to 30% less per task. Anthropic attributes the savings to fewer tokens, not a lower per-token price: both models are listed at $2 per million input tokens and $10 per million output tokens. Actual costs will vary with the task and settings.
Terminal-Bench is one company-reported result, not an independent measurement, and it does not establish savings on a customer’s workload. Anthropic says its benchmarks capture only part of a model’s performance and that Opus 5.5 remains stronger on complex, open-ended work requiring sustained judgment.
Sonnet 5.5 is available through Anthropic’s API and major cloud platforms. Developers updating existing integrations should check the migration guide: requests that explicitly disable thinking are no longer accepted, and Anthropic documents between_tools for avoiding up-front thinking.
