Chinese artificial intelligence startup DeepSeek has yet to finalize a timeline for launching its R2 model, as CEO Liang Wenfeng is reportedly dissatisfied with its current performance, according to a report by The Information on Thursday that cites sources familiar with the matter.
The R2 model, intended to succeed DeepSeek’s popular R1 reasoning model, was originally slated for a May release. The new model aims to enhance coding capabilities and expand its reasoning abilities to languages beyond English, as noted in previous reports from Reuters.
In recent months, engineers at DeepSeek have been focused on refining R2, awaiting Liang’s approval to move forward with its release, The Information reports.
However, the widespread adoption of R2 may face challenges due to a shortage of Nvidia server chips in China, linked to US export restrictions, as indicated by employees from leading Chinese cloud service providers that utilize DeepSeek’s models for their enterprise clients.
An unexpected spike in demand for R2 could potentially strain these cloud providers, who require advanced Nvidia chips for operating AI models, the report highlighted.
DeepSeek has not provided an immediate response to a request for comment from Reuters regarding the situation.
The company is reportedly collaborating with various Chinese cloud firms, supplying them with technical specifications to assist in their preparations for hosting and distributing the new model from their servers, according to The Information.
Currently, most of DeepSeek’s cloud customers utilizing the R1 model are operating it with Nvidia’s H20 chips, the report added.
Export restrictions implemented by the Trump administration in April have barred Nvidia from selling its H20 chips in the Chinese market, which were the only AI processors it could legally export to the region at that time.
© Thomson Reuters 2025
(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)
