Off-peak rates will be set at 50% of peak pricing.
DeepSeek has gained a reputation for delivering AI services at significantly lower prices compared to its more expensive western counterparts, but that trend is about to change. The company is now alerting customers about impending substantial price increases with the launch of its new model. DeepSeek V4 Pro will see its API charges rise fourfold, introducing a new tiered pricing system based on peak and off-peak hours for better resource allocation.
Effective from August 16, users of the DeepSeek V4 Pro model will be charged $3.96 for every 1 million output tokens during peak hours, a sharp increase from the previous rate of $0.87. During off-peak hours, the cost will be reduced to $1.98. The DeepSeek V4 Flash model will also see a price rise, costing $1.32 per 1 million output tokens at peak times, up from $0.28, while off-peak hours will offer a rate of $0.66.
It is worth noting that the existing API pricing was initially a temporary promotional offer that was supposed to end on May 31. Although DeepSeek had announced plans to make these discounted prices a permanent feature, the decision was later reversed in favor of the price increase.
Even with the new rates, DeepSeek’s offerings are still considerably more affordable compared to many of its rivals. For instance, the Kimi K3 model from the Chinese firm Moonshot costs $15 per 1 million output tokens, while OpenAI’s most advanced model, GPT-5.6 Sol, is priced at $30. However, OpenAI’s more economical model, GPT-5.6 Luna, is available for $1.20, which is lower than the peak hour pricing for DeepSeek V4 Flash.
