AI companies have been lowering the prices of their AI models and API services in recent months, but DeepSeek appears to be bucking the trend. The Chinese-based company has announced plans to significantly increase the pricing of its API services instead.
The announcement was made through an official notice published earlier today. While DeepSeek warned that the increase would be significant, it did not disclose the new pricing or when it would take effect.
For those unfamiliar, DeepSeek has built a reputation for offering capable AI models at highly competitive prices. According to Bloomberg, DeepSeek currently charges “US$0.14 (RM0.57) per million input tokens and US$0.28 (RM1.15) per million output tokens” for its V4 Flash model, making it one of the more competitively priced offerings in the market.
Bloomberg also compared DeepSeek’s pricing with those of its Chinese and international rivals. The report first highlighted fellow Chinese AI company Moonshot, which charges US$3 (RM12.27) and US$15 (RM61.34) per million input and output tokens respectively for its Kimi K3 model. It also compared Anthropic’s Fable 5 model, which costs US$10 (RM40.89) and US$50 (RM204.46) per million input and output tokens, respectively. If the increase is as significant as DeepSeek suggests, it could diminish one of the company’s biggest competitive advantages.
Bloomberg noted that “DeepSeek’s strategy has sparked a global debate in recent days over the cost of high-performing AI tools.” Its unusually low pricing also created what the publication described as a “death zone” for competitors, making more expensive and less capable models less attractive. “It’s also raised questions about the initial public offering plans of OpenAI and Anthropic PBC because of the risks the firm poses to their profits and business models,” Bloomberg said.
Bloomberg added that this could mark an inflection point for China’s AI market, as more companies are now offering open-source AI models at competitive prices. The publication also noted that companies and investors are increasingly looking to shift away from low-cost AI offerings towards a “commercially viable business model”.
Bloomberg also reported that DeepSeek is currently in the midst of a major fundraising round while preparing for a potential initial public offering (IPO) as early as this year. If that happens, founder Liang Wenfeng will have to balance investor expectations with the company’s rapid expansion, alongside the substantial costs of building and maintaining the computing infrastructure needed to support its AI ambitions, or so the publication says.
At the time of writing, customers can obtain DeepSeek’s V4 Flash model directly from the company itself and from other service providers who host the model. While the AI is quite capable, a sudden change in pricing may push customers, particularly businesses and developers, away in favour of cheaper alternatives. That said, because of its open-source services, it may still be a viable option for some depending on the nature of the business.
