Chinese artificial intelligence startup DeepSeek has postponed the launch of its much-anticipated R2 model, with CEO Liang Wenfeng reportedly unhappy with its current performance, according to a Reuters report citing The Information.
The R2 model, intended as an advanced successor to DeepSeek’s widely used R1 reasoning model, was initially scheduled for release in May. The new model aimed to deliver stronger coding abilities and enhanced reasoning across multiple languages beyond English.
However, according to sources cited by The Information, DeepSeek engineers have been refining the model for months, awaiting final approval from Liang before any release.
Adding to the challenges, the rollout of R2 could face major hurdles due to a shortage of Nvidia server chips in China, driven by recent U.S. export restrictions. This supply crunch has raised concerns among top Chinese cloud providers, many of whom currently use Nvidia’s H20 chips to run DeepSeek’s R1 model.
DeepSeek has been in discussions with Chinese cloud firms, sharing technical specifications for potential R2 deployment once the green light is given. However, U.S. curbs imposed in April have barred Nvidia from selling even the H20 chips—the only AI processors it was legally allowed to export to China until recently.
DeepSeek did not immediately respond to Reuters’ request for comment on the delay.
Ahmedabad Plane Crash
