Anthropic Launches Claude Opus 5.5 With Slashed Token Pricing and Frontier Coding Gains

Anthropic has officially released Claude Opus 5.5, significantly lowering inference costs across its flagship frontier tier to $4.00 per million input tokens and $20.00 per million output tokens. Under default configurations, the new pricing structure cuts costs by 40% compared to Opus 5 and makes the model 60% cheaper than Claude Fable 5.1.
Anthropic noted that Opus 5.5 matches the performance of Fable 5.1 across the majority of general evaluation suites while taking the lead across Anthropic's own lineup in software engineering and professional knowledge-work benchmarks. In cross-lab evaluations, OpenAI’s GPT-6 Astra maintains an edge on AutomationBench and Terminal-Bench-Science 0.1, though Anthropic claims Opus 5.5 delivers competitive coding output at a fraction of the cost per task.
The launch highlights rapid price compression across the frontier AI landscape, with leading labs intensifying margin and unit-economic pressure on OpenAI, Google, and open-weight alternatives to increase capabilities while expanding enterprise token efficiency.