Chinese artificial intelligence startup DeepSeek is intensifying price competition in the global AI market by sharply cutting the price of its latest flagship model, V4 Pro. As the contest among high-performance AI models shifts from pure performance to cost performance, Chinese companies are increasingly taking on U.S. big tech with low-price strategies.

The South China Morning Post reported on May 24 that DeepSeek has converted a 75 percent V4 Pro price-cut policy previously applied as a promotion into its official price. V4 Pro is being assessed as a model that offers performance comparable to the latest models from major U.S. AI companies such as OpenAI and Anthropic, but at a much lower cost.

DeepSeek, based in Hangzhou, last month unveiled the V4 series, consisting of its flagship V4 Pro model and the lighter V4 Flash model. At the time, the company flagged an aggressive pricing strategy, even mentioning the possibility of further price cuts in the second half of the year.

After the price adjustment, third-party AI benchmark firm Artificial Analysis rated V4 Pro as a top-tier global model by “intelligence” obtainable per dollar. In the AI industry, prices have been rising as supplies of computing resources needed to run advanced models tighten, strengthening a shift toward prioritising cost efficiency over pure performance.

DeepSeek says the official API price for V4 Pro is as low as $0.0036 per 1 million cached input tokens and $0.87 per 1 million output tokens. Based on that, the cost of running Artificial Analysis’ “Intelligence Index” benchmark was estimated at about $268. In the same test, OpenAI’s GPT-5.5 required about 12 times higher costs, while Anthropic’s Claude Opus 4.7 required about 19 times higher costs, the analysis found.

Other Chinese companies are also gaining visibility in the cost-efficiency race. MiniMax’s M2.7 and Xiaomi’s MiMo V2.5 Pro also ranked near the top in cost-performance standings. Alibaba has also joined the price fight, recently launching a promotion that cuts the API price of its latest model, Qwen3.7 Max, by 50 percent.

At the same time, improved cost efficiency for Chinese models is lowering barriers to adopting AI. DeepSeek launched V4 Flash at the same price as V2, which it released two years ago. The company said at the V4 unveiling in April that “prices will fall sharply in the second half.” It also said there could be room for additional price cuts if Huawei Technologies’ Ascend 950PR supernode is shipped in large volumes.

The industry sees the trend as changing the competitive landscape in the AI market. The key used to be securing the best performance, but a new standard is emerging: how much performance can be delivered for the same cost. The possibility is also rising that Chinese companies, armed with low prices and rapid performance improvements, could disrupt pricing across the global AI ecosystem.