The Chinese startup DeepSeek has created a significant stir in the artificial intelligence market, surpassing OpenAI's ChatGPT in download numbers on the App Store. The "reasoning" model DeepSeek, known as R1, was developed in a record time of just two months with a budget of less than $6 million, a fraction of the costs incurred by OpenAI and Google to train comparable systems. This was made possible by the use of less powerful chips—2048 Nvidia H800 AI accelerators.
As a result, DeepSeek has managed to offer a more cost-effective solution that, according to its developers, outperforms models such as OpenAI's GPT-4 and Meta's Llama 3.1 in terms of accuracy. This event led to a drop in the stock prices of major American companies, including Nvidia and Tesla. Investors and experts believe that DeepSeek's success could change the rules of the game in the market, forcing American companies to reconsider their approaches to AI system development.
The company has not raised funds from external sources and has not made significant moves toward monetizing its models; its focus lies in research and engineering, which aligns it with early-stage DeepMind.
DeepSeek is ready to share its achievements.
ORIENT news
