The DeepSeek-V4 series API will enable peak and valley pricing starting today: the price during the peak period will be doubled, and V4-Pro will be officially commercialized. According to the official announcement of Deep Search, the new pricing of the DeepSeek-V4 series API will officially take effect from midnight today. At the same time, V4-Pro has completed the testing phase and has been fully converted to the official version of commercial services.

The DeepSeek-V4 series API will enable peak and valley pricing from today: the price during the peak period will be doubled, and V4-Pro will be officially commercialized. According to the official announcement of Deep Search, the new pricing of the DeepSeek-V4 series API will officially take effect from 0:00 today. At the same time, V4-Pro has completed the testing phase and has been fully converted to the official version of commercial services. This price adjustment cancels the previous unified billing model and adopts the industry's rare peak-staggered tiered pricing. Charges during peak hours are directly doubled compared to idle hours. Peak and valley periods and price details: The official defines daily 9:00-12:00 and 14:00-18:00 as peak periods, and the rest of the time as idle periods. The pricing during idle periods is only half of the peak period, and the billing unit is yuan/million tokens. Taking V4-Flash as an example, the input cache hits are 0.05 yuan, misses are 1.5 yuan, and outputs are 4.5 yuan during idle periods. During peak hours, the corresponding prices rise to 0.10 yuan, 3.0 yuan, and 9.0 yuan. The overall price of V4-Pro is one level higher, with input cache hits of 0.15 yuan, misses of 4.5 yuan, and outputs of 13.5 yuan during idle periods. During peak hours, they are 0.30 yuan, 9.0 yuan, and 27.0 yuan respectively. Both models support tool calling, Responses API, Anthropic compatible interface, conversation continuation and FIM completion, with a maximum context output length of 384K. The computing power logic behind off-peak scheduling The core purpose of implementing peak-valley pricing is to optimize the allocation of computing power resources and guide developers to deploy tasks off-peak. Through market-based price leverage, the platform hopes to stabilize the computing power load during peak periods, alleviate the previous service instability problem caused by the surge in request volume, and improve overall service availability. The two V4 models are natively adapted to commercial scenarios such as long text parsing, code development, agent deployment, and enterprise batch reasoning. The 384K ultra-long context can cover most production needs. For developers, scheduling non-real-time tasks to run during idle periods is expected to significantly reduce call costs without sacrificing performance. This also marks a new stage for DeepSeek to shift from seizing the market at a low price to refined operations. via AI News (author: AI Base)