DeepSeek V4-Pro official version launched with OpenAI Responses API support

DeepSeek announced the official release of DeepSeek V4-Pro following the preview period, now available through mobile app, web version, and API. The official model features significantly enhanced agent capabilities, particularly in production environments. New features include native support for OpenAI Responses API and specific integration with Codex. The V4-Pro and V4-Flash models now support three levels of intensity in reflection mode (low, high, maximum), allowing users to adjust based on task complexity. DeepSeek will implement a new pricing scheme based on peak and valley pricing starting August 17, 2026, with reduced prices during off-peak hours, encouraging better computational load distribution among users.