DeepSeek’s Models Are Getting Closer to the Standard — A Case Study of OpenAI and Google, Anthropic, and Meta

DeepSeek hosted thousands of developers and programmers on its website and app to try out the company’s new model and shared examples of its sophisticated capabilities on social media. When investors began to question the huge sums of money being spent on artificial intelligence, the shares in US tech firms fell.

DeepSeek’s technology was developed by a small research lab in China which came from one of the best-performing quantitative hedge funds. A research paper posted online in December claims that the cost of its earlier DeepSeek-V3 model was much less than the amount it would take to build similar projects. Some models of Openai cost as much as $100 million per model. The latest models from OpenAI as well as Google, Anthropic, and Meta likely cost considerably more.

According to a report, OpenAI and Microsoft are investigating whether the Chinese rival used OpenAI’s framework to integrate its models into their own. The company believes that large amounts of data were dumped through Openai developer accounts, which it believes are affiliated with DeepSeek.

The situation is rich with irony. It was OpenAI that made the biggest leaps with its GPT model, by sucking down the entire web without consent.