Anthropic released Claude Sonnet 5.5 on Monday. The new model is an upgrade to Sonnet 5, which launched in June.
The company says the mid-tier model runs more than 30% faster than the version before it. Anthropic said the model is strongest at well-scoped everyday tasks, fixing bugs, and creating documents, slides, and spreadsheets.
Pricing stays at $2 per million input tokens and $10 per million output tokens. That is the same as Sonnet 5 and half the price of Opus 5.5, Anthropic's flagship model.
Tokens are small chunks of text that an AI reads and writes. Companies bill for them by the million.
Anthropic says Sonnet 5.5 uses nearly a third fewer tokens per task than Sonnet 5. That makes it cheaper to run than its predecessor, according to the company.
Coding Scores Beat Opus 5.5
On Terminal-Bench 4.0, Sonnet 5.5 scored 70.6%. The test checks whether an AI agent can finish complex professional tasks by typing commands on its own.
Opus 5.5 scored 66.4% on the same test. Sonnet 5 scored 10.3%.
Artificial Analysis, an independent testing firm, ran its own version of the test. It found Sonnet 5.5 at 63.6%, Opus 5.5 at 59.6%, and OpenAI's GPT-6 Astra at 59.1%.
Scores also depend on the effort setting. Higher effort makes a model think longer, which can improve answers but raises costs.
Anthropic says Sonnet 5.5 at High effort matches OpenAI's GPT-6 Sol on FrontierCode. It does so for about a fifth of the cost per task, the company said.
On GDPval-AA, which grades professional work across 44 occupations, Sonnet 5.5 scored 1844. Opus 5.5 scored 1846, and GPT-6 Sol scored 1487.
OpenAI cut GPT-6 Sol's price to $2 and $10 last week. Its mid-tier GPT-5.6 Terra lists at $2 and $12, and Anthropic published no benchmarks against it.
High Token Use at Max Effort
Artificial Analysis found that Sonnet 5.5 produces a lot of text. At max effort, it wrote about 193,000 tokens per test task, the most the firm has measured.
That is roughly 60% more than Opus 5.5. The cost came to $7.60 per task, about 50% above Sonnet 5.
Anthropic says its savings claims come from lower settings. At Medium effort, the default in its apps, the company says Sonnet 5.5 beats Sonnet 5's best coding score for less than a tenth of the cost.
Artificial Analysis says High effort offers the best value. The firm tested a pre-release build with a bug, which Anthropic expects changed little or slightly understated scores.
Anthropic's own benchmark table is self-reported. The company says Opus 5.5 remains stronger at complex work that needs sustained judgment.
Claude Haiku 5.5, built for high-volume and cost-sensitive uses, is due in the coming weeks.
