Anthropic announced on Tuesday the release of its latest model, Opus 5.5. According to the company, Opus 5.5 sets a new state of the art on coding and knowledge-work benchmarks and in many informal tasks outperforms the larger Fable model. Opus is the most capable and costly tier in Anthropic’s three-tier Claude lineup; Sonnet is the mid tier and Haiku is the fastest and least expensive.
Price and speed
The new version is considerably cheaper than its predecessor: output tokens for Opus 5.5 will be billed at $20 per million tokens, down from $25 per million on the previous model. Anthropic says other pricing metrics show similar reductions. The company also reports that Opus 5.5 runs faster, reflecting a drop in the compute required to serve the model.
Changes to communication
Opus 5.5 introduces changes to how the model communicates: it is less likely to use technical jargon and more likely to surface important information at the start of its responses. Anthropic presents these changes as improvements to clarity and usability.
Release cadence and next models
The Opus 5.5 launch comes two months after Opus 5, which was released on July 24. According to Anthropic’s announcement, Sonnet 5.5 and Haiku 5.5 — the other tiers in the Claude lineup — will be released “in the coming weeks” with similar performance upgrades.
Safety measures and limitations
Anthropic said Opus 5.5’s capabilities in biology and cybersecurity are comparable to those of Mythos, and therefore its release is subject to the same safeguards used for the company’s Fable model. Those safeguards restrict uses such as discovering exploits in compiled programs or developing recognizable biological weapons.
Opus 5.5’s safety training was broadly similar to earlier versions: alignment testing and pre-release evaluation involved outside organizations including METR and Frontier Design.
Strategy on pacing capability growth
This is Anthropic’s first model release since CEO Dario Amodei expressed support for pacing frontier progress — purposely slowing capability advancement so that alignment and risk-prevention efforts can keep up. In a post earlier this month, Amodei wrote: “I have become convinced that fully addressing the risks requires even more prudence, not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up.”
Anthropic also noted it is preparing more advanced training and evaluation systems for future models, including improved security and monitoring infrastructure, and said it expects to share more details on those efforts soon.
