The company estimates typical workloads cost forty percent less than with Opus 5.
What happened
Opus 5.5 lowers input, output and cache-read prices, while Anthropic reports faster generation and stronger performance on coding and knowledge work. The release also describes expanded testing of long tasks, impossible requests and behavior outside assigned boundaries.
Why it matters
Price per token is only part of the comparison. Context reuse and the tokens needed to finish a task influence the bill, which is why the stated typical-workload reduction differs from individual token-price cuts.
The bigger picture
Our take. Measure both successful completion and total spend on representative jobs. The vendor's results suggest a useful tradeoff to investigate, while the actual benefit depends on the workload and deployment.
Source checked November 18, 2026. Company claims are attributed; paragraphs marked “Our take” are editorial interpretation.