Anthropic cuts Claude prices by 40 percent with Opus 5.5

What's new

On September 22, 2026, Anthropic launched Claude Opus 5.5, the new top model in the Claude family. The model replaces Opus 5 and, according to Anthropic, is meant to match the performance of the company's most powerful model to date, Fable 5.1, but at a far lower price. Shortly after, rival OpenAI hit back with two new, cheaper models called Sol and Luna in the GPT-6 family. It is the latest shot in the price war that has broken out between the major AI companies.

What's clever

Opus 5.5 is built for difficult tasks such as programming, data analysis and long work processes where the AI takes many steps in a row on its own (this is called "agentic" work). Anthropic says the model handles the same tasks with 40 percent fewer AI calls than its predecessor and responds 30 percent faster. According to Anthropic, one tester completed a review of 200,000 lines of code in under three hours, something that previously took more than 20 hours with Opus 5.

Cheaper or better?

Via the developer tool (the API), Opus 5.5 costs $4 per million tokens (pieces of text) for input and $20 for output, versus $5 and $25 for Opus 5, meaning about 20 percent cheaper per token. Reuse of text (caching) has been cut all the way down to $0.20 per million tokens. All in all, Anthropic says typical work tasks become up to 40 percent cheaper. Regular users with Claude Pro (about DKK 140 a month, or DKK 120 with annual billing) automatically get access to Opus 5.5 and more usage time at no extra cost. Compared with OpenAI's GPT-6 Sol, Opus 5.5 is more expensive per token (Sol costs $2 and $10), but independent measurements show that Opus 5.5 scores higher on intelligence benchmarks and beats Sol on several coding tasks at around a third of the price. Against OpenAI's expensive top model GPT-6 Astra, Opus 5.5 is typically cheaper in normal use.

What it's good at

Opus 5.5 is especially strong at programming, large code reviews and complex tasks where the AI has to work autonomously over a long time without constantly getting new instructions. That makes it an obvious fit for companies in software development, data analysis and large document projects. For individual users, the launch mainly means that Anthropic's best AI is now both faster and cheaper to use, both in the subscription and for developers who pay per use.