DeepSeek Cuts API Prices: Off-Peak Rates Halved From August 17

According to DeepSeek, the official version of DeepSeek V4 Pro was launched on August 13 and is now available across the DeepSeek app, web platform, and API. Users can access the official V4 Pro through “Expert Mode” on the app and web platform, while the API model name remains unchanged.

DeepSeek said the latest release focuses on strengthening Agent capabilities, with particular attention to performance optimization for production environments. For Code Agent tasks in public benchmark tests, DeepSeek-V4-Pro-0813 was evaluated using the DeepSeek Harness in minimal mode with the max setting, top_p=0.95, and temperature=1.0. DeepSeek noted that test results may vary when using other frameworks.

On the API side, the DeepSeek API now natively supports the OpenAI Responses API format and has been adapted for Codex. Developers can configure Codex using the one-click setup script provided by DeepSeek.

Meanwhile, the thinking modes of V4 Pro and V4 Flash now offer three intensity levels: low, high, and max. The low setting is designed for simple tasks, high is intended for everyday Agent workloads, and max targets more complex tasks. Developers can select the appropriate level based on task complexity.

With the official release of the DeepSeek V4 series, API pricing will also be updated. DeepSeek is introducing peak and off-peak pricing, with off-peak rates set at half the price of peak periods. The new pricing will take effect at 00:00 Beijing Time on August 17.