DeepSeek R2 launch stalled as CEO balks at progress, The Information reports

Chinese AI startup DeepSeek has delayed the release of its R2 model, successor to the widely popular R1, due to CEO Liang Wenfeng's dissatisfaction with the model's current performance, according to a report by The Information. Initially scheduled for a May launch, the R2 model aims to enhance coding capabilities and improve reasoning in multiple languages beyond English. Engineers at DeepSeek have continued refining R2 pending the CEO’s approval. However, the launch faces additional challenges due to a shortage of Nvidia server chips in China, exacerbated by U.S. export restrictions, particularly the ban on Nvidia’s H20 chips imposed by the Trump administration in April. These chips are essential for running advanced AI models, and many of DeepSeek’s cloud customers still rely on them for operating R1. The potential high demand for R2 could strain Chinese cloud providers already struggling with limited access to necessary hardware. DeepSeek has engaged with cloud firms, sharing technical specifications to help them prepare for the eventual hosting and distribution of R2.