DeepSeek V4 API introduces peak-hour pricing: rates double during Beijing business hours — migrate by July 24

What's new

• Peak-hour pricing: V4 token rates double during 9–12 and 14–18 Beijing time (CST)
• Off-peak: standard rates during the rest of the day and weekends
• Migration deadline: July 24, 2026 at 15:59 UTC — deepseek-chat and deepseek-reasoner aliases stop working
• New model IDs: deepseek-v4-flash (speed/cost) or deepseek-v4-pro (maximum performance)
• Context window: V4-Pro up to 512K tokens
• Base V4 pricing: 75% cheaper than previous generation (off-peak)

Why it matters

Regional dynamic pricing is new in the AI API space. European developers calling the API during morning CET hours (9–18 CST = 2–11 CET) may pay double without any code change. The July 24 deadline is critical for production systems.

How to try it

Update model ID before July 24: deepseek-v4-pro or deepseek-v4-flash