Chinese AI company DeepSeek will quadruple the price of its flagship model

DeepSeek, a Chinese artificial intelligence developer, will sharply raise prices for its flagship V4 models—in some cases, by a factor of four. The pricing of this once-bold newcomer, which caused a stir in the stock markets, will narrow the price gap with its leading competitors in the AI sector, according to Bloomberg. The publication attributes this decision to the company’s plans to go public.

The new pricing structure will take effect on August 16 and will introduce a distinction between peak and off-peak periods. For example, users will pay a maximum of $1.32 per 1 million Input tokens (units of information that a user sends to the model) during peak hours when using the more advanced Beepseek-V4-Pro model. During off-peak hours, the rate will be half that—$0.66. Currently, the cost is $0.435 per 1 million tokens. Starting August 16, the lighter V4-Flash model will cost $0.44 and $0.22, respectively.

The 1 million Output tokens that the model issues in response to a user's request will cost more. For the V4-Pro model, the cost during peak periods will be $3.96—up from $0.87 currently.

Bloomberg reports that the Chinese AI developer’s rate hike indicates it is placing greater emphasis on profitability as it prepares for an IPO. The company may file for a listing in mainland China (not in Hong Kong) as early as this year, sources told the agency.

Last week, DeepSeek warned of an upcoming price hike without specifying the exact details. The company stated that it is reviewing and adjusting its pricing policy “to allocate resources more efficiently,” according to Bloomberg. The dynamic pricing strategy is designed to encourage developers and businesses to shift their workloads to off-peak hours, the agency explains.

Despite this increase, DeepSeek’s pricing remains lower than that of some of its main competitors. By way of comparison: Chinese competitor Moonshot AI prices its Kimi K3 model at $15 per 1 million Output tokens, while Anthropic charges $50 for its Fable 5 service.

DeepSeek’s pricing policy has sparked global debate in recent weeks about the cost of high-performance AI tools, according to Bloomberg. It is so far outside the usual range that it has given rise to the concept of DeepSeek’s “death zone, ” where more expensive or less capable models will become obsolete, the agency notes. This has also cast doubt on the IPO plans of leading startups in the sector—OpenAI and Anthropic—due to risks to their profitability and business models, the agency points out.

During its first round of external funding in early June, DeepSeek was valued at $50 billion. The company is in talks with new investors regarding the next round, in which it expects to achieve a pre-money valuation of at least 480 billion yuan ($71 billion), according to Bloomberg.

DeepSeek was founded in 2023 and is owned by the hedge fund Zhejiang High-Flyer Asset Management. The startup is attracting enormous interest from potential investors, as it is among the companies at the center of China’s efforts to strengthen the country’s position in the global AI race, according to Bloomberg.

In early 2025, DeepSeek—a company that was little-known at the time— triggered a sell-off in global stock markets with the release of its open-source artificial intelligence model, which the company claimed took only two months and less than $6 million to develop.

This article was AI-translated and verified by a human editor