DeepSeek Open Platform has declared that, commencing at 12:00 PM on September 10th, it will revise the pricing structure for the V4-Flash series large model APIs. The new pricing will be as follows: for off-peak periods, input (cache hit) will cost 0.02 yuan, input (cache miss) will cost 1 yuan, and output will cost 4 yuan. During peak hours, prices will double those of off-peak times, while the pricing for V4-Pro will remain steady. Peak hours are designated as 9:00 AM to 12:00 PM and 2:00 PM to 6:00 PM, from Monday through Friday. Following this adjustment, input prices will revert to their levels prior to the increase, though output prices will remain at twice their pre-hike amounts. The actual costs incurred by users will hinge on factors such as cache hit rates, the proportion of input to output, and the extent to which tasks are scheduled during off-peak times. Previously, the platform had ignited discussions due to price hikes, with the official explanation citing the need to allocate resources. Meanwhile, outsiders speculated that the surge was due to the high cost-effectiveness of V4 Flash 0731, which resulted in a significant uptick in usage and insufficient computing power. Furthermore, DeepSeek has commenced an internal test for the V4.1 Flash intermediate version, which will be automatically taken offline on September 10th. The company stated that costs have been minimized through architectural enhancements, and if the capabilities of the V4.1 official version are significantly enhanced, along with the price reduction strategy, it could pave the way for new development opportunities.
