Anthropic launches Opus 5.5 with lower output-token pricing
Anthropic released Opus 5.5 on Tuesday, describing it as the strongest-performing model it has tested to date for coding and knowledge work. The company said the new model outperforms the larger Fable model on many benchmarks and completed informal tasks that Fable did not complete.
The launch also lowers pricing. Output tokens for Opus 5.5 cost $20 per million tokens, compared with $25 per million for Opus 5. Anthropic said other pricing metrics also decline, while the model runs faster and has lower compute requirements.
Performance and communication changes
Opus 5.5 arrives two months after Opus 5, which Anthropic released on July 24. The earlier model's pricing and capability changes provide context because Anthropic Opus 5 pricing and capability baseline established the baseline that Opus 5.5 now lowers further while adding new performance claims.
Beyond benchmark results, Anthropic said it changed how the model communicates. Opus 5.5 is less likely to rely on jargon and more likely to place important information at the beginning of its answers. Those adjustments are aimed at making outputs easier to use without changing the company’s emphasis on coding and knowledge-work performance.
Anthropic also said Sonnet 5.5 and Haiku 5.5 will be released in the coming weeks, with similar performance improvements. The announcement does not provide dates or pricing for those forthcoming models.
Safeguards and the pace of capability gains
Anthropic said Opus 5.5 has biology and cybersecurity capabilities comparable to Mythos. It is therefore subject to the same safeguards as the company’s Fable model, including limits on use for discovering exploits in compiled programs and developing recognizable biological weapons, among other activities.
Its safety training was broadly similar to that of earlier models, including alignment testing and pre-release evaluations by outside organizations such as METR and Frontier Design. Anthropic said it is preparing more advanced training and evaluation systems for future releases, including improved security and monitoring systems.
The release is Anthropic’s first since chief executive Dario Amodei endorsed pacing frontier capability progress so that alignment progress can keep up. Anthropic also argued that public policy should have a larger role as AI systems become more capable, and said it has begun building infrastructure to support that effort.
What businesses should assess
For businesses, the release makes it practical to compare Opus 5.5 with existing deployments on task quality, response speed and token costs, while retaining the safeguards and evaluation requirements appropriate for sensitive technical work.