Anthropic officially released Claude Opus 5.5 on Tuesday, the first model in the firm's Claude 5.5 family. The newest model features a major jump in agentic coding performance while using fewer tokens, and is being positioned as a more capable autonomous "worker." Anthropic also said Opus 5.5 cuts task-level costs by 40% versus Opus 5.
In what would have normally taken an engineering team weeks, Anthropic said one tester completed a 680,000-line code migration in less than a day.
"Claude Opus 5.5 is our first release since we called for pacing the frontier," the firm said in a release. "It was tested before release by external evaluators, including Frontier Design and METR. On our automated behavioral audit, the most comprehensive alignment test we run, Opus 5.5 is the strongest-performing model we’ve tested to date. It also comes with the safeguards we’ve developed for our most capable models."
The headline API pricing for Opus 5.5 is also lower than its predecessor. Anthropic is cutting the cost of input tokens from $5 to $4 per million tokens and output tokens from $25 to $20 per million tokens. That represents a 20% reduction in both input and output pricing before accounting for the additional savings from using fewer tokens to complete tasks, the firm said.
Opus 5.5 leads Anthropic’s cited benchmarks for agentic coding, including Terminal-Bench, FrontierCode and CursorBench. In one early test, the model worked autonomously for more than 18 hours across six code repositories, while other tests showed it could perform financial analysis, build Excel models and presentations and conduct research. Anthropic said 16 of 18 internal research reports met its quality threshold, compared with none for Opus 5 or Fable 5.1.
Anthropic did caution that benchmark differences are becoming less reliable at this capability level, and said the real-world gap between Opus 5.5 and Fable 5.1 is narrower than the benchmark numbers suggest.
The firm said Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks.
The debut of the Opus 5.5 comes as the AI world at large continues to deal with public blowback amid a former researcher's post that it could destroy humanity within the next 10 years.
Last week, Anthropic published an AI transparency framework as CEO Dario Amodei called for an industry-wide slowdown. On Monday, OpenAI said the U.S. should lead a global AI framework amid rising safety concerns.