Chinese artificial-intelligence startup DeepSeek said Saturday that it will make a 75% price cut on its flagship V4-Pro AI model permanent [1].
The move signals an intensifying price war among global AI providers, potentially forcing competitors to lower costs to maintain market share.
DeepSeek reduced the API price to $0.87 per million tokens [2], down from the original price of $3 per million [3]. This permanent discount replaces a temporary promotional offer that was scheduled to expire on May 31, 2024 [4].
Company officials said falling inference costs were a primary driver for the reduction. The startup is also locked in a pricing battle with major U.S. firms including OpenAI, Anthropic, and Google [5].
Some industry analysts suggest the price drop is linked to hardware availability. The Hindu reported that the cut may be connected to an increased supply of Huawei’s Ascend 950 chips [6].
The pricing strategy positions DeepSeek as a low-cost alternative to Western models. According to the South China Morning Post, the V4 model is priced 97% lower than OpenAI’s GPT-5.5 [7].
The globally available API allows developers to integrate DeepSeek's capabilities into various applications at a fraction of previous costs. This aggressive pricing targets a broader user base as the company seeks to scale its influence in the competitive AI landscape.
“DeepSeek reduced the API price to $0.87 per million tokens.”
DeepSeek's decision to permanently slash prices suggests that the cost of running large-scale AI models is dropping faster than previously expected. By leveraging localized hardware like Huawei chips and aggressive pricing, the company is attempting to commoditize AI intelligence, shifting the competition from purely technical capabilities to operational efficiency and cost-leadership.
