Chinese LLM DeepSeek AI Surpasses ChatGPT in terms of Performance

The AI Assistant from Chinese startup DeepSeek surpassed competitor ChatGPT to take the top spot among free apps on the US Apple App Store. Since its release on January 10, the artificial intelligence application has become increasingly popular among US users, thanks to the DeepSeek-V3 model, which its developers claim "tops the leaderboard among open-source models and rivals the most advanced closed-source models globally," according to app data research firm Sensor Tower.
 

The milestone demonstrates how DeepSeek has profoundly impacted Silicon Valley, challenging conventional wisdom regarding US dominance in AI and the efficacy of Washington's export restrictions aimed at China's cutting-edge chip and AI capabilities. Advanced chips are needed to power the training of AI models, ranging from ChatGPT to DeepSeek. Since 2021, the Biden administration has expanded the breadth of prohibitions intended to prevent these chips from being transferred to China and used to train AI models for Chinese companies.

However, according to a report published last month by DeepSeek researchers, the DeepSeek-V3 spent less than US$6 million on training using Nvidia's H800 CPUs.

The assertion that the chips utilized were not as powerful as the most cutting-edge Nvidia products Washington has aimed to keep out of China, along with the comparatively low cost of training, has led US tech executives to doubt the efficacy of tech export regulations, even though this detail has since been refuted.

The business behind DeepSeek, a little startup based in Hangzhou that was established in 2023 when search engine giant Baidu unveiled the first Chinese AI large-language model, is not well known.  Although numerous Chinese IT companies, both big and small, have since developed their own AI models, DeepSeek is the first to receive praise from the US tech industry for matching or even outperforming state-of-the-art US models.