Half price during idle hours, doubled during peak hours: DeepSeek API peak and valley pricing is online, with V4Pro increasing by up to 11 times. Deep Search will officially launch the new DeepSeek API peak and valley time-of-use pricing rules from 0:00 on August 17. The API call price during idle periods is only half of that during peak periods, aiming to guide developers through market-based price leverage for off-peak scheduling.

Half price during idle hours, doubled during peak hours: DeepSeek API peak and valley pricing is online, with V4Pro increasing up to 11 times. Deep Search officially launched the new DeepSeek API peak and valley time-of-use pricing rules from 0:00 on August 17. The price of API calls during idle hours is only half of that during peak hours. It aims to use market-oriented price levers to guide developers to schedule large model tasks in off-peak hours to alleviate daytime computing power congestion and improve platform operation stability. Officially, 9:00 to 12:00 and 14:00 to 18:00 Beijing time are the peak hours of computing power every day, and the rest of the time is the idle low-peak period. The billing unit is yuan/million tokens. The latest quotations of the two main models are effective simultaneously: the input (cache hit/miss) and output prices of DeepSeek V4Flash during peak hours are 0.10 yuan, 3.0 yuan, and 9.0 yuan respectively, and the corresponding half price during idle periods; the three prices of DeepSeek V4Pro during peak hours are respectively 0.30 yuan, 9.0 yuan, and 27.0 yuan, and during idle periods they are 0.15 yuan, 4.5 yuan, and 13.5 yuan respectively. Compared with the initial pricing when the official version of DeepSeek V4Pro was launched on August 12 - input (cache miss) 3 yuan, output 6 yuan, cache hit input as low as 0.025 yuan, under the new regulations, the cache hit input price increased by 1100% during peak hours, the output price increased by 350%, and the cache miss input increased by 200%. Even if you choose to call during idle periods, the three prices are 6 times, 1.5 times and 2.25 times the original price respectively. The implementation of the peak and valley pricing mechanism marks DeepSeek's official farewell to the previous "extreme cost-effectiveness" pricing strategy, and provides a new reference paradigm for the refined operation of domestic AI large model platforms' computing power. It is worth noting that even at the peak price after the price adjustment, the output unit price of V4-Pro is still only about one-third of that of the Claude flagship model, and it is about one-twenty-fifth of it in idle time. via AI News (author: AI Base)