DeepSeek V4 API Price Increase: Peak and Off-Peak Rates

Peak hours, off-peak hours, and a single flat rate that no longer exists

DeepSeek said on August 13, 2026 that prices for developer access to its V4-Flash and V4-Pro models would rise by 50 percent to more than 1,100 percent depending on the model, token type, and time of use, with the new rates taking effect at 16:00 UTC on August 16. The increase replaces a single flat rate with peak and off-peak billing: peak hours are 01:00 to 04:00 UTC and 06:00 to 10:00 UTC, and off-peak rates are set at half the peak rate, according to DeepSeek's pricing page. [1]

V4-Flash output tokens rise to $1.32 per million at peak and $0.66 per million off-peak, up from a flat $0.28 per million. V4-Pro output tokens rise to $3.96 per million at peak and $1.98 per million off-peak, up from a flat $0.87 per million. Input pricing moved too: V4-Flash cache-miss input tokens rise to $0.44 per million at peak, up from $0.14, and V4-Pro cache-miss input tokens rise to $1.32 per million at peak, up from $0.435. [1]

DeepSeek said the change was meant "to allocate resources more reasonably," framing the peak and off-peak split as a way to push developer workloads toward less congested hours. DeepSeek's own API documentation confirms the model naming behind the increase: the deepseek-v4-pro endpoint was updated to DeepSeek-V4-Pro-0813, with the calling method unchanged. [1][3]

Atlas interpretation: Even the new peak rates leave DeepSeek well below frontier Western pricing. Quartz cites Fortune's reporting that Anthropic's Fable 5 charges $50 per million output tokens, more than twelve times DeepSeek's new V4-Pro peak rate. The increase is best read as a retreat from an unsustainable floor rather than a move toward parity: DeepSeek is still selling access far below what comparable frontier labs charge, it is just no longer selling it at cost during the hours everyone wants it. [1]

The vague warning, then the number

DeepSeek had signaled the increase a week earlier. On August 6, 2026, one week after releasing DeepSeek-V4-Flash-0731, the company said a "significant" price hike was coming and told users to "plan their usage accordingly," without giving figures or a date. The South China Morning Post reported the pressure came from surging global demand for the company's low-cost models, which had strained its ability to keep the aggressive pricing in place. [2]

The warning drew criticism from developers before any numbers were public, with one telling the Post: "DeepSeek choosing to raise prices at this time, isn't this just asking for trouble?" given competition from newer American models offering comparable performance at comparable prices. [2]

Atlas interpretation: The increase reverses a pricing trajectory DeepSeek had been running since it shipped V4 in April 2026 with open weights. Six weeks after that release the lab made its promotional discount on V4-Pro permanent rather than letting it expire. Quartz notes DeepSeek had also offered a 75 percent promotional discount on V4-Pro through May 5, which underscores how far the pricing has now moved in the other direction inside a few months. [1]