Anthropic released Claude Opus 5.5 with prices of $4 per million input tokens and $20 per million output tokens, both 20% below Opus 5, and reduced computing requirements. Typical workloads cost 40% less at default settings in Anthropic's tests, while output generation is more than 30% faster. The model matches Claude Fable 5.1 on most tasks. Cache reads now cost $0.20 per million tokens, a 60% reduction.

Opus 5.5 is now the default in Claude Code and the Claude app, including Cowork, for Pro, Max and Team subscribers. Claude Code's five-hour session limits increased 20%, while the model's lower pricing separately lets usage limits go 25% further. Subscribers also receive a usage reset they can apply through Oct. 22. The model is available in services including GitHub Copilot, Cursor and Perplexity Computer. Sonnet 5.5 and Haiku 5.5 will follow in the coming weeks.

In Box's enterprise knowledge work tests, Opus 5.5 used 63% fewer tokens and ran 30% faster than Opus 5. It also topped the Artificial Analysis Intelligence Index with a score of 58 at maximum reasoning effort. Savings varied by setting and workload, however: in that evaluation, its cost per task was level with Opus 5 because it generated about 1.6 times as many output tokens.