China’s DeepSeek AI Startup Unveils New Model in Challenge to OpenAI

Chinese AI startup DeepSeek has released a new version of its large language model, DeepSeek-V3-0324, as part of its growing efforts to compete with established US-based companies like OpenAI.

This new model represents a significant update to the company’s V3 series, with notable improvements in areas such as reasoning and coding performance, according to benchmark tests published on AI development platform Hugging Face.

DeepSeek has attracted attention in the AI community, particularly in Silicon Valley, due to its ability to deliver a model with capabilities comparable to OpenAI’s ChatGPT, but at a fraction of the development cost. The company claims its V3 model was trained with less than $6 million worth of computing power, utilizing 2,000 Nvidia H800 chips. Despite these cost-effective training methods, DeepSeek’s model is said to perform on par with some of the most advanced AI models developed by major US firms like OpenAI and Meta.

The release of DeepSeek’s new model adds to its increasing presence in the competitive global AI market. The company launched its V3 model in December and followed up with the R1 model in January, demonstrating rapid progress in its capabilities.

While the emergence of DeepSeek has raised questions about the billions of dollars that US tech companies have invested in advanced chips and large data centers to train their AI models, it also represents a shift in the global AI race. Once dominated by US companies, the competition is now intensifying with China’s growing involvement in the AI sector. This shift has sparked comparisons to the Cold War-era “Sputnik moment,” when the US first realized the challenge posed by advancements from its geopolitical rivals.

With input from FOX Business and Reuters.