Anthropic has released Claude Opus 5.5, its latest model for coding and knowledge work, with higher performance, lower prices and faster output than Opus 5. The company says the model reaches state-of-the-art results on several evaluations and outperforms the larger Claude Fable model on a number of benchmarks and informal tasks.
Opus sits above Sonnet and Haiku in Anthropic’s standard three-tier Claude lineup, while Fable represents a larger class of model. Anthropic released Opus 5.5 just two months after Claude Opus 5 arrived on July 24.
Opus 5.5 Costs Less and Runs Faster
According to Anthropic’s Opus 5.5 announcement, output costs $20 per million tokens, down from $25 for Opus 5. Input tokens cost $4 per million, while Anthropic says typical workloads should cost about 40% less overall.
The model also generates output more than 30% faster than Opus 5 and requires less computing power to serve. Anthropic has changed its communication style as well, with Opus 5.5 designed to use less jargon, place important information earlier in responses and make longer interactions easier to follow.
Sonnet 5.5 and Haiku 5.5 are expected to follow in the coming weeks. Anthropic says the two models will receive similar improvements in performance, efficiency and safety.
Anthropic Applies Stronger Safeguards to Opus 5.5
Anthropic says Opus 5.5 approaches Claude Mythos in biology and cybersecurity capabilities, placing it within a higher-risk category than previous Opus models. The company is therefore applying safeguards similar to those used for Fable, including restrictions on some cybersecurity activities and biological work that could have harmful applications.
Anthropic previously detailed how its biology safeguards restrict potentially dangerous dual-use requests while allowing more routine educational and professional tasks. Opus 5.5 also underwent alignment testing and external pre-release evaluations from organizations including METR and Frontier Design.
The release is Anthropic’s first new model since CEO Dario Amodei called for AI developers to pace advances in frontier capabilities. Amodei argued that capability development should proceed at a rate that gives safety research, alignment work and outside evaluation enough time to keep up.
Anthropic said it is already preparing more advanced evaluation, security and monitoring systems for future models. The company also said public policy should take a larger role as AI systems become more capable and that it plans to share more details about those efforts.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.