Anthropic announced Claude Opus 5.5 on September 22, 2026. Its reported 40% saving versus Opus 5 concerns typical workloads at default settings in company tests; it is not a universal discount. Standard API input and output prices are $4 and $20 per million tokens, each 20% lower than Opus 5. Fast mode costs $8 and $40 respectively. These are API rates, not subscription fees. The announcement lists availability on the Claude Platform and major cloud platforms.
For a practical comparison, choose one recurring task, such as turning approved meeting notes into an action list. Keep the input and acceptance criteria identical. Record the model, mode, total charge, retries, and time spent checking the result. Include failed attempts when comparing the cost of an acceptable output. This is our suggested workflow, not an Anthropic benchmark.
A lower token price alone cannot establish your total savings. Review accuracy and omissions before adopting the result, and check the current mode and billing terms before expanding the test.