Early this morning, the Claude Opus 4.1 model was released, Pro/Max/Team users can use it on the web, and the API is open.Claude.ai is now available with enhanced performance and the same price.
Comprehensive improvement in capabilities
According to Anthropic's official statement, Claude Opus 4.1 is a direct replacement for Opus 4. It has been comprehensively improved in terms of programming, writing, tool calling, and Agent.
• Agent capabilities: can accurately handle long-term tasks and complex enterprise workflows, and performs well in TAU-bench;
• Advanced programming: can complete long-distance programming tasks with thousands of steps, and supports a single 32k text output;
• Stronger search: can independently conduct research tasks for several hours, and conduct internal and external information search more effectively;
• Stronger writing: write more like people
According to public benchmark test data, Claude Opus4.1 has surpassed the old version in every list, as shown in the figure below.
Programming related benchmarks
Let’s talk specifically about the SWE-bench list, which is a programming list.
In the list, Opus 4.1 has the highest score of 74.5%, followed by Claude Sonnet 4. The following figure is more specific data
SWE-bench
appendix
Appendix 1: Claude Opus 4.1 Information Card/Technical Report
https://assets.anthropic.com/m/4c024b86c698d3d4/original/Claude-4-1-System-Card.pdf Attachment 2: Related links
Claude Webapp
www.claude.ai
Claude Opus official information
https://www.anthropic.com/claude/opus