Anthropic ships Claude Sonnet 5.5 at Sonnet 5 prices

Anthropic released Claude Sonnet 5.5 on 28 September 2026, the second model in the Claude 5.5 family. The company says it generates output more than 30 percent faster than Sonnet 5 and costs up to 30 percent less for most work. List pricing stays at 2 US dollars per million input tokens and 10 US dollars per million output tokens; the savings are meant to come from using fewer tokens per task. On the API the model is called claude-sonnet-5-5, and it is also available on Amazon Web Services, Google Cloud and Microsoft Azure. According to Simon Willison, Sonnet 5.5 is now also the model behind the free tier of claude.ai.

On the agentic benchmark Terminal-Bench 4.0, Sonnet 5.5 scores 70.6 percent according to Anthropic, against 10.3 percent for Sonnet 5. Anyone running Sonnet with thinking switched off has to move to the new between_tools setting before upgrading: it suppresses thinking at the start but still allows it between tool calls. Willison also reports that at the highest effort level, "max", the model used 128,000 tokens in his test and stopped without producing a result. He had seen the same failure earlier with Opus 5.5.

Why it matters: For automations built on Sonnet 5, the switch is essentially a change of model ID at the same list price. The exception is pipelines without thinking, which need to move to between_tools first. Whether the promised savings materialise only shows once token usage is compared on real workloads, especially since the price round at OpenAI and Anthropic already shifted the numbers last week. Teams using the "max" level should cap its token budget until a fix arrives.