Anthropic has released Claude Opus 5.5, a new AI model the company says performs at roughly the same level as its Claude Fable 5.1 model on most work, but at a significantly lower cost.
The model, announced on Tuesday, Sept. 22, is the first in Anthropic’s new Claude 5.5 family. According to the company’s announcement, Opus 5.5 costs about 40 percent less to run on typical workloads than Opus 5, which launched in July. Anthropic said Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks.
Opus 5.5 is also the first new Anthropic model since CEO Dario Amodei called on AI companies to “pace the frontier” and put the brakes on AI development for safety reasons.
Claude Opus 5.5 leads in coding and knowledge work, Anthropic says
Anthropic said Opus 5.5 leads its benchmark suite in agentic coding, computer use, and knowledge work. On GDPval-AA v2.1, which tests real-world work across 44 occupations, the company reported that Opus 5.5 scored 1,846. Fable 5.1 scored 1,735 and Opus 5 scored 1,708.
Credit: Anthropic / Mashable screenshot
On Terminal-Bench 4.0, Anthropic reported Opus 5.5 at 66.4 percent, compared with 57.9 percent for OpenAI’s GPT-6 Astra, which launched earlier this month. Astra still led on some tests. Those included Terminal-Bench-Science 0.1, on which it scored 64.6 percent, compared to Opus 5.5’s 58.7 percent.
Anthropic also cautioned that benchmark margins have become a less reliable guide at this level of capability. The company said the real-world gap between Opus 5.5 and Fable 5.1 is narrower than the scores suggest.
The company also said the model writes more clearly than Opus 5, which it described as a common complaint about the earlier model.
Mashable Light Speed
Want to learn more about getting the best out of your tech? Sign up for Mashable’s Top Stories and Deals newsletters today.
Claude Opus 5.5 safety and safeguards
As the company’s first new model since Amodei’s “pacing the frontier” essay, safeguards are a particular concern.
The company said outside evaluators, including METR and Frontier Design, tested the model before release. Anthropic also said Opus 5.5 posted the best scores of any model to date on its automated behavioral audit.
This approach differed from Opus 5’s. Mashable reported in July that Anthropic had deliberately left cybersecurity tasks out of that model’s training. Anthropic now says Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity. Because of that, it is shipping with safeguards similar to those in Fable 5.1. Most cybersecurity tasks will be rerouted to Opus 4.8, and vetted organizations can apply for broader access to biology through a new Life Sciences Verification Program.
Anthropic acknowledged limits to its testing. The company said it has seen signs that Opus 5.5 often suspects it is under evaluation, making its real-world behavior harder to assess. OpenAI has reported similar behavior with GPT-Astra.
How to try Claude Opus 5.5
Claude Opus 5.5 is available now across Anthropic’s platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can access it through the API as claude-opus-5-5.
API pricing is:
-
$4 per million input tokens
-
$20 per million output tokens
-
$0.20 per million tokens for cache reads
That’s 20 percent less than Opus 5 for input and output tokens, and 60 percent less for cache reads, according to Anthropic. For comparison, GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens.
Anthropic also says it’s raising five-hour usage limits for Pro, Max, and Team subscribers.
Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
Topics
Artificial Intelligence
Anthropic
