Claude Opus 5.5 delivers Fable 5.1 performance – and costs 40% less

Anthropic released Claude Opus 5.5, less than two months after Claude Opus 5. The company says Opus 5.5 delivers Fable 5.1 performance for most work and costs about 40% less to run, with tokens priced 20% lower than with Opus 5. Anthropic says the model generates output more than 30% faster than Opus 5 and needs fewer tokens for higher quality work.

Anthropic says it will also launch Claude Sonnet 5.5 and Haiku 5.5 over the coming weeks. Opus 5.5 is available today.

For subscription users, Anthropic is raising five-hour usage limits by 20%. The company says both five-hour and weekly usage limits go further because Opus 5.5 costs less than Opus 5.

Anthropic says Opus 5.5 communicates more naturally than prior models. It describes the model as its strongest performing to date on alignment testing, citing improvements on behaviors it links to recent cybersecurity incidents, including biased reasoning and attempting to escape a sandbox. The company says it used external testing providers and a similar class of safeguards to Fable 5.1 on cybersecurity, biology, and frontier LLM development. If safeguards fire, requests fall back from Opus 5.5 to Opus 4.8; most cybersecurity tasks will be rerouted to Opus 4.8, and biology and LLM development requests to Opus 5. Vetted organizations can apply to Anthropic's Life Sciences Verification Program, and those approved through its Cyber Verification Program can start using Opus 5.5 in a few weeks.

Customer evaluations were cited. Yashodha Bhavnani, VP of AI Products at Box, said Claude Opus 5.5 used a third of the tokens Opus 5 did, with answers 40% less verbose without losing accuracy. GitHub chief product officer Mario Rodriguez said that in testing across GitHub Copilot CLI and VS Code, Opus 5.5 used among the fewest tokens and steps measured, and in VS Code solved more terminal tasks than Opus 5 in less than half the steps. Deloitte Consulting CIO Carl Bennett said that at its lowest effort setting, Opus 5.5 caught 72% of known bugs in code reviews versus Opus 5's 56% at high effort, with fewer false alarms. John Ruelas, a staff software engineer at Ramp, said Opus 5.5 follows the company's writing rules and that a design spec came out usable with minimal edits.