Anthropic released the Claude Sonnet 5.5 model on September 28, 2026. The company claims the new model operates over 30% faster and costs up to 30% less per task compared to its predecessor, Sonnet 5. Pricing remains unchanged at $2 per million input tokens and $10 per million output tokens.
The model is designed for everyday tasks, including bug fixing and document creation. It includes cyber safeguards and anti-distillation classifiers to prevent the extraction of reasoning data for training rival systems. Anthropic previously launched Opus 5.5 on September 22, 2026.
Benchmarking firm Artificial Analysis assigned Sonnet 5.5 a score of 56 on its Intelligence Index, placing it second behind Opus 5.5. The firm noted that at maximum effort, Sonnet 5.5 used approximately 193 000 output tokens per task, which is 60% higher than Opus 5.5.
In contrast, Balyasny Asset Management reported different results on its internal finance tasks. The firm stated that Sonnet 5.5 used about 121 000 tokens per answer, compared to 497 000 tokens required by Sonnet 5. Anthropic noted that the initial tests by Artificial Analysis involved a pre-release build with a bug.
