Claude Opus 5.5 lands: Fable-level coding at 40% lower cost

Anthropic just shipped Claude Opus 5.5, and the headline number is price, not power. It performs about as well as Fable 5.1 on most tasks and costs roughly 40% less to run than Opus 5. API pricing drops to $4 per million input tokens and $20 per million output tokens, down from $5 and $25, and output comes about 30% faster.

The benchmarks are strong too. It scores 66.4% on Terminal-Bench 4.0 (Fable 5.1 gets 55.8%) and 81.8% on OSWorld 2.0 for computer use. Anthropic says it beats GPT-6 Astra on FrontierCode at about 20% of the cost per task. It also targets "Claudish" writing: the model puts the key point first, uses less jargon and follows style instructions more closely.

It's also the first release since Anthropic's call to pace the frontier. Opus 5.5 gets the best score so far on the company's behavioral audit and makes 85% fewer attempts to get around its boundaries than Opus 5. It's live now on the API as claude-opus-5-5 and on AWS, Google Cloud and Azure. Sonnet 5.5 and Haiku 5.5 are due in the coming weeks.

Why it matters: the gains here are in price and reliability, not raw capability, and that makes top-tier agentic coding far cheaper to run at scale.