China's DeepSeek Slashes AI Prices With 75% Discount on V4-Pro Model
DeepSeek recently introduced its V4 series in preview, which includes two versions: the more powerful V4-Pro and a lighter, more affordable Flash variant.
China’s DeepSeek is not just launching a new AI model. It is aggressively cutting prices to make sure people actually use it. The company is offering a 75% discount on its newly unveiled DeepSeek-V4-Pro model for developers until May 5, while also slashing API costs across its lineup, including reducing input cache pricing to just one-tenth of previous levels.
A New Model, But Pricing Is the Real Story
DeepSeek recently introduced its V4 series in preview, which includes two versions: the more powerful V4-Pro and a lighter, more affordable Flash variant. The Pro model is designed for high-performance workloads and has been adapted to run on Huawei’s AI chips, reflecting China’s broader push toward reducing reliance on foreign hardware.
In terms of benchmarks, DeepSeek claims the Pro version performs better than other open-source models on world knowledge tasks, trailing only Google’s closed-source Gemini-Pro-3.1.
So yes, the model is competitive. But that’s not what makes it interesting.
Advertisement
This Is a Pricing War
The 75% discount is not a one-off promotional gimmick. It is part of a larger strategy.
DeepSeek has already been known for pushing down AI costs, and this move doubles down on that approach. By cutting both model pricing and API costs simultaneously, the company is making it significantly cheaper for developers to build and deploy AI applications.
Advertisement
In practical terms, this lowers the barrier to entry. If running AI models becomes cheaper, more developers experiment, more startups build products, and more enterprises consider switching providers.
Built for AI Agents, Not Just Chatbots
DeepSeek says the V4 models are particularly suited for “AI agents,” which are systems capable of performing more complex, multi-step tasks rather than just answering questions.
These workloads require more computing power, which is exactly why pricing becomes critical. Lower costs make these advanced use cases more viable at scale.
Huawei Chips Are a Strategic Shift
Another important detail is hardware. The V4 models are designed to work with Huawei’s chip ecosystem instead of relying on Nvidia hardware. This is a direct response to export restrictions and reflects a broader effort by China to build a self-sufficient AI stack.