DeepSeek implements new peak and off-peak API pricing

DeepSeek has implemented a new tiered pricing structure for its API services, effective August 17, 2026. The company has moved away from flat rates toward a peak and off-peak model designed to manage server utilization and inference costs.

Under the new schedule, peak hours are defined as 9:00–12:00 and 14:00–18:00 Beijing Time. During these windows, rates for the V4 Pro model are higher, while off-peak rates are set at half the peak price. For example, V4 Pro input tokens with a cache miss cost 9 yuan per million during peak hours, compared to 4.5 yuan during off-peak hours. The most significant change is seen in cache-hit input prices, which saw an increase of up to 1100%.

This shift reflects a broader industry trend where AI providers are utilizing time-of-day differentials, cache-aware discounts, and model tiers to optimize capacity. The adjustment specifically targets developers using API services and does not affect standard users of the DeepSeek web or mobile applications.