DeepSeek is overhauling how it prices access to its flagship models, introducing time-of-day billing that will raise some rates by as much as twelvefold, according to the company’s official API pricing documentation.
Starting at 16:00 UTC on August 16, 2026, the DeepSeek API will charge different rates for DeepSeek-V4-Pro and DeepSeek-V4-Flash depending on when a request is made. Peak hours run 01:00–04:00 and 06:00–10:00 UTC — spanning the middle of the workday in China — with every other hour billed at a lower off-peak rate.
How much prices are rising
The steepest increase hits V4-Pro’s cached input tokens, which jump from a flat $0.003625 per million to $0.044 per million at peak — roughly 12 times higher, and still 6 times higher during off-peak hours. Output tokens on the same model climb from $0.87 to $3.96 per million at peak, or $1.98 off-peak. V4-Flash, the smaller and cheaper of the two models, sees steep but less dramatic increases: peak output pricing rises to $1.32 per million tokens, up from a previous flat rate of $0.28.
Part of a broader pattern
The change arrives four days after V4-Pro left preview and became generally available, and follows more than a year in which Chinese AI labs have competed with US rivals largely on price. DeepSeek says the tiered structure is meant to spread developer demand more evenly across the day rather than concentrating requests during its busiest hours — a tactic already common among cloud and inference providers managing capacity constraints.
Even with the increase, several outlets covering the change note that DeepSeek’s API pricing remains below many frontier US models’ published rates, though the gap narrows sharply during peak hours under the new structure. The Chinese lab rose to prominence in early 2025 on the strength of aggressively low pricing, so this marks a notable shift in strategy for a company that built its reputation as the cheap alternative.