Anthropic releases Opus 5.5 with lower prices and Fable-level performance

Anthropic announced Opus 5.5, a new top-tier Claude model the company says raises state-of-the-art performance for coding and knowledge work while reducing latency and serving costs. The release, made two months after Opus 5, reportedly outperforms the larger Fable model on several benchmarks and informal tasks and introduces communication-style changes to be clearer and less jargon-heavy.

By AI Newsroom· Reviewed by Pranav, Founder & Editor-in-ChiefPublished about 1 hour agoUpdated about 1 hour ago0 views

Anthropic releases Opus 5.5 with lower prices and Fable-level performance

Why It Matters

The update combines higher capability with lower run costs, which could affect developer adoption and competitive positioning among large-model providers; it also brings Opus into the same safety-restricted treatment as Anthropic's Fable and Mythos-level capabilities. Anthropic frames the launch alongside ongoing work to slow the pace of frontier capability growth and strengthen safety and evaluation processes.

Key Facts

Model released: Opus 5.5 released Tuesday (two months after Opus 5, which launched on July 24).
Position in lineup: Opus is the most capable and most expensive tier in Anthropic's Claude lineup (Sonnet middle tier, Haiku fastest/cheapest).
Performance claim: Anthropic says Opus 5.5 outpaces the larger Fable model on many benchmarks and succeeded on informal tasks Fable failed to complete.
Pricing change: Output tokens charged at $20 per million tokens for Opus 5.5, down from $25 for the previous model; other pricing metrics also fell.
Latency/compute: Anthropic reports Opus 5.5 is faster to run and requires less compute to serve.

Anthropic on Tuesday unveiled Opus 5.5, an upgrade to its top-tier Claude model that the company says improves performance on coding and knowledge-work benchmarks while cutting latency and serving costs. The company positions Opus at the high end of its three-tier Claude family, with Sonnet in the middle and Haiku as the lower-cost, faster option. According to Anthropic, Opus 5.5 surpasses the larger Fable model on a number of benchmarks and handled several informal tasks that Fable did not complete. A notable economic change with the release is a reduction in output-token pricing: Anthropic said it will charge $20 per million output tokens for Opus 5.5, down from $25 under the prior model. The company also reported lower compute requirements and faster runtime for the new model, and indicated parallel price drops across other usage metrics. Sonnet 5.5 and Haiku 5.5 are slated to arrive in the coming weeks with comparable improvements. Opus 5.5 introduces changes to how the model communicates, with the company saying it now avoids jargon more often and places key information earlier in responses. On safety, Anthropic said Opus 5.5 is comparable to its Mythos model in biological and cybersecurity capabilities, and therefore is subject to the same restrictions that govern Fable—limiting use cases like discovering exploits in compiled programs or developing recognizable biological weapons. The model’s safety training followed approaches similar to previous releases and included alignment testing and pre-release evaluations by external organizations such as METR and Frontier Design. The launch also follows an internal shift at Anthropic: CEO Dario Amodei recently endorsed pacing frontier capability growth to better align progress in capability with advances in alignment and risk prevention. The company said it is preparing more advanced training and evaluation systems for future models, along with improved security and monitoring infrastructure, and expects to disclose additional details on those efforts in time.