Anthropic has launched its latest AI model, Opus 5.5, which the company describes as its most capable system to date. The model demonstrates performance levels exceeding the larger Fable model in various benchmarks and specific tasks. It also features updated communication styles, prioritizing key information and reducing technical jargon.
The company has reduced the cost of using the model, with output tokens now priced at $20 per million tokens, down from $25 for the previous version. The model also requires less compute power, resulting in faster execution speeds. These updates follow the release of Opus 5, which debuted on July 24, 2026.
Anthropic plans to release Sonnet 5.5 and Haiku 5.5 in the coming weeks. Opus 5.5 is subject to the same safety protocols as the Fable model due to its advanced capabilities in cybersecurity and biology. These safeguards restrict the model's ability to assist in developing biological weapons or discovering software exploits.
This release marks the first model launch since CEO Dario Amodei announced a strategy to pace the development of AI capabilities to align with safety progress. The model underwent alignment testing and external evaluations by organizations such as METR and Frontier Design, with plans for more advanced security systems in future iterations.
