Chinese AI firm DeepSeek has announced a 75% permanent price reduction for its flagship V4-Pro AI model, slashing costs to a quarter of their previous levels. The company revealed on May 23 that API pricing for the V4-Pro model would now vary between 0.025 yuan and 6 yuan per million tokens, depending on usage type. This represents a significant drop from the previous rates ranging between 0.1 yuan and 24 yuan per million tokens. Tokens denote units of text processed by AI systems.
The exact reason for the price cut was not explicitly stated by DeepSeek, raising questions about a potential correlation with the increased availability of Huawei’s Ascend 950 chips. These chips have been instrumental in enhancing the performance of DeepSeek’s V4 models.
This pricing adjustment aligns with the evolving landscape of China’s semiconductor and AI sectors, which have been impacted by US export restrictions on advanced chip sales to China. These restrictions have restricted NVIDIA’s ability to market some of its cutting-edge processors in China, thereby opening doors for domestic chip manufacturers like Huawei.
Upon the launch of the V4 model last month, DeepSeek highlighted that the Pro version could be priced significantly higher than the less powerful Flash version due to limitations in high-end computing capacity. The company also hinted at the possibility of substantial price reductions once Huawei’s Ascend 950 supernodes are widely available in the latter half of the year, as reported by Reuters.
