Anthropic Ships Claude Opus 5.5, Claiming Fable 5.1-Level Output at 40% Lower Cost

Anthropic has released Claude Opus 5.5, the first model in a new generation, which the company says matches Claude Fable 5.1 on most tasks while running at roughly 40 percent lower cost than Opus 5. The launch pushes the Opus line into a slot Anthropic has not occupied before: a flagship-tier model priced closer to its mid-range siblings. According to Anthropic's own benchmarks, Opus 5.5 also outperforms OpenAI's GPT-6 Astra on most of the tasks tested, even though it costs considerably less per run.

The pricing gap is the headline number. A 40 percent reduction against Opus 5 changes the math for teams running large inference workloads, where per-token cost compounds quickly across agents and batch jobs. Anthropic is positioning the model as a direct swap rather than a downgrade, which means customers who held back from Opus 5 on budget grounds now have a reason to re-evaluate. The company has not yet published a full per-token price sheet alongside the announcement.

On capability, Anthropic claims parity with Fable 5.1 across most workloads, a model that until now sat at the top of the company's stack. The benchmark comparison against GPT-6 Astra covers unspecified tasks, so the results should be read as vendor-reported rather than independently verified. Still, a cheaper model landing ahead of a rival's flagship on most tests is a claim OpenAI will face pressure to answer, and the two labs have been trading benchmark leads throughout the year.

Anthropic is also promising less "Claudish" writing — a reference to the recognizable tics that have become shorthand for AI-generated prose. That detail matters less for benchmark charts than for the growing set of customers using Claude to draft emails, reports, and marketing copy, where stylistic sameness has become a real complaint. Whether the change is measurable or mostly a positioning line will surface once users put it through long-form writing tasks.

The rest of the generation is coming. Sonnet 5.5 and Haiku 5.5 are expected within weeks, which would give Anthropic a refreshed lineup at three price points. That cadence suggests the company is trying to compress the gap between its top model and its cheaper ones, rather than sell capability as a premium tier. Competitors will be watching how quickly the cheaper models inherit Opus 5.5's gains.

The open question is verification. Anthropic's numbers are self-reported, and independent evaluations of Opus 5.5 against GPT-6 Astra and Fable 5.1 have not yet appeared. Buyers deciding whether to migrate production traffic will want third-party results on cost-per-task, not just benchmark scores, before committing. Anthropic has not said when it will publish detailed pricing or a technical report.