DeepSeek V3 — $5.6M Training Cost Shatters Foundation Model Economics

In one sentence DeepSeek V3: 685B MoE model trained for $5.6M that outperforms GPT-4o and Claude 3.5 Sonnet on coding and math. MIT license. Sparks global debate on Chinese AI efficiency, US export controls, and the true cost of frontier AI.

Needs review Reputable source

ShareLinkedIn X

A Chinese AI lab called DeepSeek released a massive open model called V3 that shocked the entire AI industry. The model outperforms OpenAI's GPT-4o and Anthropic's Claude 3.5 Sonnet on coding and mathematics — yet it was trained for only 5.6 million dollars. For comparison, the leading US models are believed to cost hundreds of millions to train. This single fact upended a widely held assumption: that only companies with enormous capital could produce frontier AI. The model is released under an MIT license, meaning anyone can use it commercially for free. The release triggered immediate reactions globally — US policymakers questioned whether export controls on AI chips were actually working, investors reconsidered the valuations of US AI companies, and developers rushed to run the model locally. It was the most disruptive open-model release since the original Llama, and arguably more consequential.