Anthropic refined its flagship model. Opus 4.1 achieves better results not only in programming

On Tuesday, Anthropic unveiled Claude Opus 4.1, an updated version of its flagship artificial intelligence model. The new variant brings improvements in programming, task automation, and logical reasoning. The original Opus 4, along with the Sonnet 4 model, was introduced in May of this year.

Better results in programming and analysis

Claude Opus 4.1 achieved 74.5% success on the SWE-bench Verified benchmark, which tests models’ capabilities in real-world programming tasks. The model also shows improvements in detailed data analysis and information retrieval using automated tools.

According to GitHub, the new version improves performance across most functions compared to the previous Opus 4, with the most significant progress noted in refactoring code across multiple files. Japanese firm Rakuten Group states that Opus 4.1 can accurately identify necessary fixes in large codebases without unnecessary modifications or introducing bugs.

Extended reasoning capabilities

Claude Opus 4.1 is among the hybrid reasoning models, which combine standard responses with extended thinking up to 64,000 tokens. This feature allows the model to analyze complex problems in more detail before formulating a response.

Windsurf, a software development company, reports a one standard deviation improvement over Opus 4 in its benchmark for junior developers. This improvement corresponds to a similar leap as the transition from the Sonnet 3.7 model to Sonnet 4.

Availability and pricing policy

Claude Opus 4.1 is now available for paying users of the Claude service and in the Claude Code application. Developers can use the model via API under the name claude-opus-4-1-20250805, and it is also available on Amazon Bedrock and Google Cloud’s Vertex AI platforms.

Anthropic refined its flagship model. Opus 4.1 achieves better results not only in programming Adam Kurfürst News Adam Kurfürst News

Anthropic maintains the same pricing structure as the previous Opus 4, so users do not have to anticipate higher costs when upgrading to the new version. The company recommends upgrading from Opus 4 to version 4.1 for all uses.

Anthropic also announces that in the coming weeks, it plans significantly larger improvements to its models than what the current Opus 4.1 update brings. However, it did not specify them further.

Do you use any of the Claude models in your work?

Source: Anthropic