PANews reported on May 23 that Anthropic officially released two new models, Claude Opus 4 and Claude Sonnet 4, at the developer conference. Opus 4 performed best on the SWE-bench verification set (72.5%, reaching 79.4% in high-computing mode), becoming the world's leading automatic programming model. Sonnet 4 also reached 72.7%, surpassing OpenAI o3 and Codex-1. Rakuten tests showed that Opus 4 can continuously program for 7 hours, stably handling complex tasks and breaking industry records. The new models support parallel tool usage and improved memory mechanisms, and Claude Code is now fully open.
