Anthropic unveils Claude Opus 5.5 — on par with Fable 5.1 and 40% cheaper

Anthropic has introduced Claude Opus 5.5, the first model in the new Claude 5.5 family. According to the company, it performs at the level of Claude Fable 5.1 on most tasks but costs about 40% less than Opus 5.

Tokens cost $4 and $20 per million for input and output, versus $5 and $25 for the previous model. Cache reads were reduced separately — $0.20 versus $0.50 per million: those account for the bulk of spending on agentic and coding tasks. The model also generates responses more than 30% faster.

On the Terminal-Bench 4.0 benchmark, Opus 5.5 scores 66.4% versus 57.9% for GPT-6 Astra. Anthropic itself cautions, though, that at this level of capability score differences no longer predict real-world differences between models well, and the gap to Fable 5.1 is smaller than the numbers suggest.

Early testers report a noticeable gain on the hardest tasks: one of them migrated 680,000 lines of code in under a day — work that would have taken a team weeks.

On the automated behavior audit, which checks nearly 2,000 scenarios, Opus 5.5 gives the best result among Anthropic's models. It takes hard-to-reverse actions less often and is more resistant to prompt injection; for biology and cybersecurity the same restrictions as Fable 5.1 apply, with some requests routed to other models.

The model is already available on all platforms, including AWS, Google Cloud, and Microsoft Azure. Verified organizations can apply to the Life Sciences Verification Program, and the Cyber Verification Program will be expanded.

Sonnet 5.5 and Haiku 5.5 are promised for release "in the coming weeks" — with some of the same performance, efficiency, and safety improvements.

Image:

Translation editor: Салтанов А.