DeepSeek宣布DeepSeek-V4-Pro正式版上线[2]，在App、网页端和API平台同步更新[2]。新版本增强了Agent能力，原生支持OpenAI Responses API格式，针对性适配Codex[2]，并新增低、高、最大三档思考强度设置供用户选择[2]。
同时，DeepSeek推出更新的API定价策略[1]，引入峰值和非峰值费率机制，其中非峰值费率较峰值费率低50%[1][2]，以支持更灵活的工作负载调度[1]。新定价机制将于北京时间2026年8月17日0时起生效[2]。
DeepSeek has released its V4-Pro model with significant updates to API pricing and functionality [1][2]. The new model became available across the company's app, web platform, and API simultaneously [2]. Starting August 17, 2026 at midnight Beijing time, DeepSeek will implement a peak and off-peak pricing structure, with off-peak rates set at 50% of peak rates [1][2]. This tiered approach is designed to enable more flexible workload scheduling for users [1].
The V4-Pro release introduces several technical enhancements beyond pricing adjustments [2]. The model now natively supports OpenAI's Responses API format and includes targeted optimization for Codex [2]. Users can access the new capabilities through "Expert Mode" in the app or web interface [2]. Additionally, the model offers three configurable thinking intensity levels—low, high, and max—allowing developers to balance response depth with latency requirements [2].
