Anthropic introduced Claude Opus 5.5 on September 22. The company says it approaches Claude Fable 5.1 on most work, surpasses it on some coding evaluations, and delivers its strongest Opus performance yet in behavioral auditing and safety.
Standard API pricing is $4 per million input tokens and $20 per million output tokens. Anthropic estimates that typical token-billed workloads cost about 40% less than with Opus 5. The base input and output prices themselves are 20% lower; cheaper cache reads contribute to the larger overall saving.
For developers running long coding sessions, that distinction matters: the final bill depends on how much context is reused, how many output tokens the agent generates, and whether a faster processing mode is enabled. The quoted prices are for standard processing.
The competition among frontier coding models keeps getting tighter. A model that produces reliable code at a lower total cost can be useful even when another model leads a particular benchmark.
Official information: Anthropic's Claude Opus overview.
Adapted from the original Chinese article, published on September 23, 2026.
Comments NOTHING