DeepSeek V4 official release mid-July, introduces peak pricing

DeepSeek V4 official release is set for mid-July. The Flash variant has been the world's most-called API model for six consecutive weeks. New pricing introduces peak-hour surcharging: API prices double during Beijing time 9AM-12PM and 2PM-6PM. Maximum output is extended to 384K tokens, and the 1M context window is now standard. Legacy model IDs deepseek-chat and deepseek-reasoner will be retired on July 24 — all applications must migrate before then.