Anthropic released Opus 5.5 on Tuesday, cutting the price of output tokens to $20 per million from $25 and saying the model performs better than its predecessor on coding and knowledge work. The company said the new version runs faster, writes more plainly, and puts important information earlier in messages.
The launch matters because Opus is Anthropic’s top tier, the model users turn to when they want the company’s most capable system. Anthropic said Opus 5.5 set a new state of the art in coding and knowledge work performance and outpaced Fable in many benchmarks, while also succeeding in informal tasks that Fable failed to complete. That gives the release immediate weight for users of Anthropic’s top-end Claude model, especially with the company signaling that Sonnet 5.5 and Haiku 5.5 are coming in the next few weeks.
The timing also lands just about two months after Opus 5 arrived on July 24 and only days after Dario Amodei said the industry needed more prudence. Earlier this month, he said fully addressing the risks requires more caution and that capabilities advancement should be paced so risk prevention has time to keep up. Opus 5.5 is the first model release since that shift was made explicit, which makes the pricing cut and performance gains feel less like a sprint and more like a recalibration.
Anthropic is trying to sell speed and restraint at the same time. The company says Opus 5.5 is comparable to Mythos in biology and cybersecurity capabilities, but it is still subject to the same safeguards as Fable, including limits on discovering exploits in compiled programs and developing recognizable biological weapons. Its safety training was broadly similar to earlier models, and it also went through alignment testing and pre-release evaluation by outside organizations including METR and Frontier Design. That combination suggests Anthropic is moving ahead on capability while keeping the guardrails in place.
The next test is whether the coming Sonnet 5.5 and Haiku 5.5 releases carry the same pattern: lower cost, sharper performance, and tighter control. If they do, Anthropic will have shown that it can keep pushing the frontier without abandoning the slower pace Amodei has argued for, even as the gap between speed and safety remains the central question around its most advanced models.
