DeepSeek Launches V4 Pro into General Availability with New Pricing Structure

DeepSeek announced the full launch of its V4 Pro model, making it accessible through its web platform, Expert Mode, and API. The model now supports the OpenAI Responses API and has been integrated with Codex, facilitating easier migration of existing workflows. The company introduced three levels of reasoning effort—low, high, and max—to cater to different use cases. Starting August 16, DeepSeek implemented new pricing with peak and off-peak rates, where off-peak costs are set at 50% of peak rates. These changes indicate a maturing monetization strategy and a competitive focus on enhancing the quality of AI agent systems.