Anthropic has launched Claude Sonnet 5.5, an updated mid-tier model that reportedly operates over 30% faster than its predecessor. The company maintains pricing at $2 per million input tokens and $10 per million output tokens, which is half the cost of the Opus 5.5 model.
In coding performance, Sonnet 5.5 achieved a 70.6% success rate on the Terminal-Bench 4.0 test, outperforming Opus 5.5 at 66.4% and the previous Sonnet 5 at 10.3%. Independent testing by Artificial Analysis confirmed these results, placing the model ahead of competitors like OpenAI's GPT-6 Astra.
The model's efficiency depends on user-defined effort settings. While high-effort configurations increase token usage and costs, Anthropic claims that medium-effort settings allow Sonnet 5.5 to surpass the coding capabilities of Sonnet 5 at less than a tenth of the expense.
On the GDPval-AA benchmark, which evaluates professional tasks across 44 occupations, Sonnet 5.5 scored 1844, nearly matching the 1846 score of Opus 5.5. Anthropic noted that Opus 5.5 remains superior for complex tasks requiring sustained judgment.
Anthropic also announced that the Claude Haiku 5.5 model, designed for high-volume applications, is scheduled for release in the coming weeks. The company continues to compete with OpenAI, which recently adjusted pricing for its GPT-6 Sol and GPT-5.6 Terra models.
