At the recent developer conference, Anthropic introduced two groundbreaking programming models: Claude Opus 4 and Claude Sonnet 4. These models showcased impressive capabilities during testing, with Claude Opus 4 achieving a top score of 72.5% on the SWE-bench validation set and reaching an impressive 79.4% in high-computing mode, solidifying its position as a leading global automatic programming model. Claude Sonnet 4 also performed exceptionally well, scoring 72.7%, surpassing OpenAI’s o3 and Codex-1 models in benchmarks conducted by Rakuten. These new models are not only powerful but efficient as well; Claude Opus 4 can program continuously for seven hours while handling complex tasks with ease, setting a new industry benchmark. This advancement is also fueled by improved memory mechanisms within the models, leading to the greater accessibility of Claude Code via enhanced tools and interface.