Anthropic has announced the release of its Claude Opus 4.7 model, habr.com reported. This model represents a significant step forward in programming: in real-world task tests, Cursor scores rose from 58% to 70%, while Rakuten noted a threefold increase in solved production tasks compared to Opus 4.6.
Among the technical changes, the support for images up to 2576 pixels, improved file memory handling in long sessions, and the addition of a new xhigh effort level are noteworthy. Additionally, the /ultrareview command was introduced for identifying bugs and architectural issues in code. The model's pricing remains unchanged: 5 dollars per million input tokens (approximately 64,500 soums) and 25 dollars per million output tokens (approximately 322,500 soums).
Read “Zamin” on Telegram!
