Anthropic has launched Claude Opus 4.1, the latest version of its AI model, for paid users, API clients, and via Amazon Bedrock and Google Cloud’s Vertex AI.

The company confirmed that pricing for Opus 4.1 remains the same as the previous version.

According to Anthropic, Opus 4.1 achieves a 74.5% score on the SWE-bench Verified coding benchmark, and offers improvements in research and data analysis.

User feedback from GitHub, Rakuten, and Windsurf highlights better performance in code refactoring and debugging.

Anthropic said it will roll out more updates to its models in the coming weeks.

🔗 Source: Anthropic

🧠 Food for thought

1️⃣ AI coding models show incremental but meaningful performance gains

Claude Opus 4.1’s benchmark performance demonstrates the pattern of steady improvement in AI coding capabilities.

The model achieved 74.5% on SWE-bench Verified, representing a 2 percentage point improvement over Claude 4’s 72.5% score12.

While this may seem modest, such incremental gains translate to significant real-world impact when applied across thousands of coding tasks.

This improvement pattern reflects the broader evolution of AI coding tools, where each iteration builds upon previous capabilities rather than delivering revolutionary leaps.

The specific focus on multi-file code refactoring and debugging precision suggests AI models are tackling increasingly complex, real-world development scenarios that require understanding relationships across entire codebases2.

2️⃣ Enterprise validation drives AI coding tool development

The involvement of major companies like Rakuten Group and GitHub in providing specific feedback on Claude Opus 4.1 highlights how enterprise adoption is shaping AI coding tool evolution.

Rakuten Group specifically praised the model’s ability to “pinpoint exact corrections within large codebases without making unnecessary adjustments or introducing bugs”2.

Recent Claude developments