On 22 September, Anthropic launched Opus 5.5. According to the company, it comes close to Fable 5.1 on many work tasks and is better at keeping answers clear and useful.
Anthropic also describes a lower typical cost than Opus 5 when using the API. This concerns developers using the model and should not be confused with a lower monthly price for all ordinary subscriptions.
For Pro and Max users, the company highlights higher usage limits. Sonnet 5.5 and Haiku 5.5 are expected in the coming weeks, but that remains a plan, rather than meaning the models are already available on everyone’s account.
The interesting question is how these improvements will feel in ordinary tasks. A model that asks a good follow-up question or explains a change clearly can be more helpful than one that simply writes longer answers.