Anthropic Debuts Opus 5.5 With Lower Prices

Anthropic has officially launched Opus 5.5, the latest update to the most capable tier of its Claude AI lineup. Released on Tuesday, just two months after the July 24 debut of Opus 5, the new model establishes a new state-of-the-art for coding and knowledge work. It is also significantly cheaper and faster than its predecessor, with output tokens now priced at $20 per million tokens compared to the previous rate of $25. This price drop is accompanied by a reduction in the compute power required to run the model, translating to faster processing times for users.

In terms of performance, Opus 5.5 outpaces Anthropic's larger Fable model across multiple benchmarks and successfully completes informal tasks that Fable could not resolve. Beyond raw capability, the update refines how the model communicates. Opus 5.5 is designed to write with less jargon and structure its responses to place critical information right at the beginning of its messages. Practitioners can expect more direct, readable outputs that streamline integration into daily workflows.

Because Opus 5.5 matches the Mythos model in its biology and cybersecurity capabilities, Anthropic has applied the same strict safety protocols used for Fable. These guardrails prevent the model from being used to find vulnerabilities in compiled software or assist in creating biological weapons. The release represents Anthropic's first launch since CEO Dario Amodei advocated for pacing frontier AI development to ensure alignment keeps up. External groups like METR and Frontier Design evaluated the model prior to its release. Looking ahead, Anthropic plans to launch Sonnet 5.5 and Haiku 5.5 in the coming weeks with similar performance upgrades.