Anthropic releases Opus 5.5 at a lower price
The new top-tier Claude model is positioned as faster and more capable for coding and knowledge work, while keeping Anthropic’s existing safety limits.
Anthropic has released Opus 5.5, the newest and most capable model in its Claude lineup, and says it sets a new benchmark for coding and knowledge-work tasks. The launch stands out not just for the performance claims, but because Anthropic says the model is also cheaper and faster to run than the previous version. At the same time, the company is keeping Opus 5.5 under the same safety restrictions used for its most sensitive AI systems.
The release lands at a time when major AI companies are under pressure to improve performance without raising risks. Opus 5.5 now sits at the top of Anthropic’s three-tier Claude family, above Sonnet and Haiku, and the company is also signaling that it wants advanced AI to be more affordable to deploy.
Anthropic says Opus 5.5 outperforms larger Fable
According to Anthropic, Opus 5.5 delivers state-of-the-art results in coding and knowledge work. The company also says it outperforms the larger Fable model in many benchmarks and completed several informal tasks that Fable could not.
That comparison helps explain why the launch drew attention. In AI, a bigger model does not always mean better real-world results, so claims that a smaller or more efficient system can match or beat a larger one tend to matter for businesses and developers following the market.
Anthropic is not presenting this as a dramatic redesign of its product line. Instead, it is framing Opus 5.5 as a stronger version of its top-tier model for the same kinds of tasks people already use AI for — writing code, handling information, and supporting knowledge work.
Lower prices come with faster responses
Opus 5.5 is also cheaper to use than its predecessor. Anthropic says output tokens will cost $20 per million tokens, down from $25 for the previous model. The company says other pricing metrics have also been reduced.
Anthropic also says the model is faster to run. As IT-PUB News notes, that could matter for companies using AI at scale and trying to keep costs under control, since lower serving demands can make a model easier to deploy more broadly.
For users and businesses, this may be the most practical part of the announcement. Better performance matters, but so does whether an advanced model can be used without pushing costs higher. Lower serving costs can make a model more appealing for products that rely on frequent AI calls, though Anthropic did not give details on specific commercial use cases.
Opus 5.5 is designed to write more plainly
Anthropic says Opus 5.5 also changes how it communicates. The company says the model is less likely to use jargon and more likely to put important information at the start of its replies.
That may sound minor, but it has real weight for everyday use. AI systems are judged not only by whether they answer correctly, but by whether they explain things clearly enough to be useful. A model that gets to the point faster and avoids overly technical language may simply be easier to work with for people outside engineering.
Anthropic did not say this shift was aimed at a particular audience. Still, the update suggests the company is paying attention to how people read AI-generated text, not just how models perform on benchmarks.
Existing safety limits remain for sensitive uses
Anthropic says Opus 5.5 is comparable to Mythos in its biology and cybersecurity capabilities, which means it is subject to the same safeguards as the company’s Fable model. Those safeguards limit how the models can be used to discover exploits in compiled programs or to develop recognizable biological weapons, among other tasks.
This is one of the more significant parts of the release. More capable models can also bring more concerns about misuse, especially in cybersecurity and biology-related areas, and Anthropic is explicitly tying capability gains to risk controls.
The company did not say these safeguards are new. Instead, it said the existing protections remain in place as it rolls out a stronger version of its top model. That makes the launch notable not just as a product update, but as another example of how AI firms are trying to widen access while keeping restrictions on sensitive capabilities.
The release comes after Amodei called for slower frontier AI
Opus 5.5 is Anthropic’s first model release since CEO Dario Amodei said he supports pacing the frontier — deliberately slowing the advance of AI capabilities so alignment work can keep up.
In a post earlier this month, Amodei wrote that fully addressing the risks requires “even more prudence,” including not only investment in risk prevention but also a slower rate of capabilities progress so safety efforts have time to catch up.
That gives this launch extra context. Anthropic is still releasing a new top-tier model, but it is doing so while publicly stressing caution. The company is not presenting Opus 5.5 as a push for maximum capability at any cost. It is pairing the release with safeguards and pointing to future safety infrastructure as part of the same approach.
Anthropic says more safety updates are on the way
Anthropic said the safety training for Opus 5.5 was broadly similar to that of its predecessors, including alignment testing and pre-release evaluation by outside organizations such as METR and Frontier Design. At the same time, the company said more advanced training and evaluation systems are already being prepared for future models.
Those future systems include improved security and monitoring tools, according to the announcement. Anthropic also said it expects to share more details about those efforts soon.
The message around Opus 5.5 is not just about a model getting better. Anthropic is also using the launch to show where it sees the next stage going — stronger performance and lower pricing now, with more formal safety infrastructure still being built around future releases.
