Anthropic releases Claude Sonnet 5.5 at the same token price

Anthropic has released Claude Sonnet 5.5, a model aimed at everyday work such as fixing bugs, writing documentation, making presentations and building spreadsheets. It is available now through Anthropic and major cloud platforms, including Amazon Web Services, Google Cloud and Microsoft Azure.

The company says Sonnet 5.5 generates answers more than 30% faster than its predecessor and uses fewer tokens—the units of text an AI model processes—so tasks can cost up to 30% less. Its listed price per million tokens is unchanged from Sonnet 5: $2 for input and $10 for output. Anthropic’s cost savings are estimates; independent testing has not yet confirmed them.

How close is it to Opus?

On GDPval-AA, a test covering workplace tasks across 44 professions, Sonnet 5.5 scored 1,844 points. Opus 5.5 scored 1,846, while the older Sonnet 5 scored 1,449. On CursorBench, a test based on coding sessions, Sonnet 5.5 scored 55.5%, two points below Opus 5.5.

Coding results improved sharply over Sonnet 5 on some measures. Sonnet 5.5 scored 70.6% on Terminal-Bench 4.0, compared with 10.3% for its predecessor. There is a caveat: on FrontierCode, it scored lower at its highest effort setting, called “Max,” than at “Xhigh.” Anthropic says a code-review process sometimes caused timeouts or changes beyond the assigned task.

Available with extra safeguards

Anthropic says Sonnet 5.5 is more capable at cybersecurity tasks than Sonnet 5. High-risk requests are routed to Sonnet 5, while qualified professionals can apply for access to higher-risk capabilities through an expanded verification program. The model also includes safeguards intended to deter attempts to copy its outputs to train other systems.

Sonnet sits below Anthropic’s Opus model, which is intended for more complex work. Anthropic has also announced Haiku 5.5, a lower-cost model for high-volume use, but says it will arrive in the coming weeks.