Anthropic released Claude Sonnet 5.5, which scores 56 on the Artificial Analysis Intelligence Index, placing it second only to Opus 5.5. The model demonstrates the highest output tokens per task performance observed, with an 18-point improvement over its predecessor.