DeepSeek is raising the API prices for its latest AI models by more than fourfold as it introduces peak and off-peak pricing.
From August 16, DeepSeek V4 Pro will cost $3.96 per one million output tokens during peak hours, up from $0.87. The price will fall to $1.98 during off-peak hours.
DeepSeek V4 Flash will cost $1.32 per one million output tokens at peak times, up from $0.28, and $0.66 during off-peak hours.
The company said the new structure was intended to allocate resources more reasonably. Off-peak pricing will be set at half the peak-hour rate.
DeepSeek gained attention by offering AI services at prices below many Western competitors. Its existing API prices had initially been presented as a promotion due to end on May 31.
The company later said the discounted rates would become permanent, but has now decided to proceed with higher prices for its V4 models.
Even after the increases, DeepSeek V4 Pro will remain cheaper than some competing models. Kimi K3, made by Chinese company Moonshot, costs $15 per one million output tokens, while GPT-5.6 Sol costs $30.
However, DeepSeek V4 Flash will cost more at peak hours than GPT-5.6 Luna, which is priced at $1.20 per one million output tokens.
