DeepSeek has introduced a permanent 75% discount on its V4-Pro AI model

DeepSeek, a Chinese artificial intelligence developer, has announced a permanent 75% reduction in the cost of using its V4-Pro language model, setting the price at one-quarter of the previous rate.

This was reported by Reuters.

The company has reduced the API cost for the model from the previous range of 0.1–24 yuan to 0.025–6 yuan per million tokens, making access to the technology significantly cheaper.

DeepSeek notes that this decision is long-term in nature and effectively establishes a new lower price floor in the generative AI market. As a result, the Chinese model becomes significantly cheaper than its counterparts from OpenAI and Anthropic, intensifying global competition in the AI sector.

The company does not specify whether the decision is linked to an increase in the supply of Huawei Ascend 950 chips, which are used to optimize the system’s performance. At the same time, DeepSeek had previously warned of a shortage of computing power, which caused the V4-Pro version to cost significantly more than the simplified version.

Initially, the company planned to lower prices only after the large-scale deployment of Huawei’s infrastructure in the second half of the year, but it implemented this move earlier.

Amid U.S. sanctions restricting the supply of advanced Nvidia chips to China, Huawei is actively expanding domestic production and sales of its solutions. At the same time, restrictions on chip manufacturing equipment remain a key factor slowing the scaling of Ascend production.

Thus, DeepSeek is not only lowering the cost of its products but also intensifying pressure on the global AI solutions market, shifting the competitive balance between the U.S. and China.