Tech war: US firms embrace DeepSeek’s AI model despite scrutiny
From Nvidia to Microsoft and Amazon, US companies have rushed to adopt the Chinese start-up’s R1 model
Top chip designer Nvidia has made DeepSeek’s R1 model available to users of its NIM microservice since Thursday, saying the model provides “state-of-the-art reasoning capabilities”, “high inference efficiency”, as well as “leading accuracy” for tasks requiring logical inference, reasoning, mathematics, coding and language understanding.
DeepSeek’s open-source reasoning model, R1, released on January 20, has shown capabilities comparable to OpenAI’s closed-source GPT models in certain areas, but at significantly lower training costs.
Smaller US tech firms have also embraced DeepSeek’s models, with AI search start-up Perplexity offering uncensored results from R1 on topics including the Tiananmen Square crackdown in 1989, which are blocked by the original model.