Anthropic launches cheaper, higher-performing Opus 5.5 model

Anthropic on Tuesday released Opus 5.5, a new AI model the company says sets a new benchmark for coding and knowledge-work performance. According to Anthropic, the model outperforms its larger sibling, Fable, on numerous benchmarks and completed several informal tasks that Fable could not.

Lower prices, faster performance

The new model is also notably cheaper to run than its predecessor: output tokens are priced at $20 per million, down from $25 for Opus 5. Other pricing metrics have dropped by similar margins. Opus 5.5 also runs faster, reflecting a broader reduction in the computing resources needed to serve it. Anthropic also revised how the model communicates, making it less likely to use jargon and more likely to lead its responses with the most important information.

The release arrives just two months after Opus 5 launched on July 24. Anthropic says Sonnet 5.5 and Haiku 5.5 will follow "in the coming weeks" with comparable performance gains.

Safeguards and a slower pace of development

Anthropic says Opus 5.5's biology and cybersecurity capabilities are comparable to those of its Mythos model, meaning it is subject to the same safeguards applied to Fable — restrictions that limit its use in discovering exploits in compiled software or developing recognizable biological weapons.

This is the first Anthropic release since CEO Dario Amodei said the company would deliberately slow the pace of capability advances to let safety work keep up. Safety training for Opus 5.5 broadly followed the same process as prior models, including alignment testing and evaluation by outside organizations such as METR and Frontier Design. Anthropic said it is already preparing more advanced training and monitoring systems for future models, and expects to share further details soon.