DeepSeek V4 API pricing officially announced with peak/valley mechanism
DeepSeek announced V4 API pricing on August 5, with V4-Flash output as low as ¥2 per 1M tokens and V4-Pro doubling prices during peak hours.
On August 5, DeepSeek officially announced the full pricing scheme for its V4 API series, introducing the industry's first peak-valley pricing mechanism. The V4-Flash version is priced at ¥2 per 1M output tokens during off-peak hours, while the V4-Pro version is ¥6 per 1M output tokens. International pricing for V4-Flash is $0.14 input and $0.28 output per 1M tokens.
The peak-valley mechanism specifies that from 9:00 AM to 12:00 PM and 2:00 PM to 6:00 PM Beijing time on weekdays, both V4-Pro and V4-Flash API prices double. During peak hours, V4-Flash is ¥2 input and ¥4 output per 1M tokens, while V4-Pro is ¥6 input and ¥12 output per 1M tokens. Off-peak prices remain unchanged, and cached-input prices stay low regardless of timing.
A Haitong International research report published on August 5 noted that V4-Flash's average cost per benchmark test is just 3 cents, far below Anthropic Claude Fable 5's $3.15 and OpenAI GPT-5.6 Sol's $1.86. Low-cost models will accelerate Agent and AI application penetration, with application companies refocusing on customer acquisition, industry data, and workflow loops.