Anthropic released its newest artificial intelligence model, Opus 5.5, on Tuesday. The company states the release sets a new state-of-the-art benchmark for coding and knowledge work performance.
Opus is the top tier in Anthropic's three-tier Claude lineup. Sonnet sits in the middle, and Haiku operates as the fastest and cheapest option.
Anthropic reports that Opus 5.5 outpaces the larger Fable model in multiple benchmarks. The new version also succeeded in several informal tasks where Fable failed.
The new model costs significantly less than its predecessor. Output tokens cost $20 per million tokens for Opus 5.5, down from $25 for the previous model. Other metrics show similar price drops, and the model runs faster due to lower compute requirements.
Opus 5.5 also changes communication habits. The model uses less jargon and places important information at the beginning of its messages.
The launch occurs two months after the release of Opus 5 on July 24. Anthropic announced that Sonnet 5.5 and Haiku 5.5 will launch in the coming weeks with comparable performance gains.
Opus 5.5 matches Mythos in biology and cybersecurity capabilities. Because of this, the release falls under the same safeguards used for the Fable model. These safeguards restrict usage related to discovering exploits in compiled programs and developing biological weapons.
This release follows a strategic shift by Anthropic CEO Dario Amodei. Earlier this month, Amodei supported calls to pace frontier AI development to match progress in safety alignment.
External organizations like METR and Frontier Design conducted pre-release evaluation and alignment testing for Opus 5.5. Anthropic is developing advanced training and evaluation systems for future releases.
Anthropic stated that Sonnet 5.5 and Haiku 5.5 will arrive in the coming weeks.
