China’s DeepSeek V4 And Qwen Reshape The Open-Source AI Race

Source Domain: www.forbes.com

Illustration of DeepSeek AI on January 29, 2025. DeepSeek AI is a Chinese artificial intelligence company and the name of its conversational agent, which exploits a large language model. In early 2025, DeepSeek unveils a new model published in open source under the MIT license. Artificial intelligence. (Photo by Riccardo Milani / Hans Lucas / Hans Lucas via AFP) (Photo by RICCARDO MILANI/Hans Lucas/AFP via Getty Images)

Hans Lucas/AFP via Getty Images

Over the weekend, DeepSeek did something that put the entire US AI press into motion. The Chinese AI lab released V4-Pro and V4-Flash on April 24 under an MIT open-source license. Two days later, on Saturday, it announced a 75% promotional discount on V4-Pro. Less than 24 hours after that, on Sunday, it cut input cache hit prices to one-tenth of what they had already been.

The numbers, when stacked, are genuinely arresting. V4-Pro now lists at $1.74 per million input tokens and $3.48 per million output tokens. With the promotional discount applied, output drops to roughly $0.87 per million. Compare that to GPT-5.5 at $30.21 per million output tokens, or Claude Opus 4.7 at $25. By that math, DeepSeek is shipping output that is 35 times cheaper than GPT-5.5 and 29 times cheaper than Opus 4.7.

A senior DeepSeek researcher posted on X late Sunday night: “Go wild and have fun.” A few hours later, another X user replied: “This is not normal. Every AI company is out here chasing profits. Why does DeepSeek keep dropping prices this hard?”

That last question is the right one. But the answer it leads to is not the one most coverage is providing.

The DeepSeek price war is real. The story being told about what it means is wrong on at least three different levels. Each of those levels matters separately, and together they paint a much more interesting picture than the panic narrative suggests.