DeepSeek raises V4 API prices as demand pressures capacity

DeepSeek is increasing API prices for its V4 model family, ending part of the low-cost advantage that helped distinguish the Chinese AI vendor. Some rates are rising by more than 1,100%, though the impact depends heavily on whether workloads run during peak or off-peak periods. The new rates take effect across most regions on August 16.

V4-Flash and V4-Pro now use a pricing structure that separates peak and half-price off-peak usage. Flash is positioned for volume workloads, while Pro is aimed at more complex tasks. Analysts said the headline increases look steep, but DeepSeek’s cache discounts and the fact that 17 of every 24 hours remain half price could allow buyers to limit higher costs by shifting schedulable work.

The move reflects broader pressure on AI infrastructure as demand rises faster than available compute. Developers are expected to feel the increases most directly, while enterprises may rely more on model routing, workload scheduling, and compatibility across providers to control spending. Analysts said the larger issue is whether foundation model vendors can keep charging premiums as comparable capabilities become available through multiple technical and commercial paths.