Anthropic's Claude Sonnet 5.5 closes the gap with Opus

Anthropic has launched Claude Sonnet 5.5, the second model in its Claude 5.5 family and an upgrade to Sonnet 5 released earlier this year.

The new model generates output more than 30% faster than its predecessor and typically uses fewer tokens to complete the same work. Anthropic says that can make individual tasks up to 30% cheaper, although API pricing remains unchanged at $2 per million input tokens and $10 per million output tokens. Cache reads cost $0.20 per million tokens.

Some of the largest gains are in coding. Sonnet 5.5 scored 70.6% on the agentic coding benchmark Terminal-Bench 4.0, up from 10.3% for Sonnet 5. On CursorBench 4.0, which evaluates coding agents using tasks drawn from real Cursor sessions, the new model scored 55.5%, compared with 34.1% for its predecessor.

Sonnet 5.5 also comes close to the more expensive Opus 5.5 on some professional tasks. The two models scored 1,844 and 1,846 respectively on GDPval-AA v2.1, a benchmark covering real-world work across 44 occupations. Anthropic says Opus 5.5 remains clearly stronger for complex, open-ended work that requires sustained judgment.

The release also marks the first time Anthropic has launched a Sonnet model with the cybersecurity safeguards used for its most capable models. Sonnet 5.5 also includes protections designed to prevent attackers from extracting the model's reasoning through distillation attacks.

Claude Sonnet 5.5 is available across Anthropic's platforms, as well as through AWS, Google Cloud and Microsoft Azure. Claude Haiku 5.5, the smallest model in the family, is expected to follow in the coming weeks.