(Bloomberg) — DeepSeek has reopened access to its core programming interface after nearly a three-week suspension, resuming a service key to wider adoption of an AI model that’s proven remarkably popular since its emergence last month.
Most Read from Bloomberg
The 20-month-old Chinese startup, which stunned Silicon Valley and markets in January with an AI platform that rivals OpenAI’s, said it’s again allowing customers to top up credits for use on its application programming interface. DeepSeek suspended top-ups in early February because of capacity shortages. While those have now resumed, server resources will remain constrained during the daytime, a DeepSeek representative said in a verified company group chat on WeChat.
DeepSeek resumed top-ups the same day that Alibaba Group Holding Ltd. launched a preview of its latest model, QwQ-Max, underscoring the deepening competition within China’s nascent AI industry. TikTok owner ByteDance Ltd. also said it’s testing a self-developed deep reasoning model with a limited set of users, adding the prospect of another enhanced DeepSeek rival.
Alibaba pledged this week to invest $53 billion over three years to bolster its cloud computing and AI infrastructure, in a major pivot for the e-commerce pioneer. On Tuesday, the company declared plans to open-source QwQ-Max.
DeepSeek’s arrival reinvigorated the Chinese tech scene and triggered a rally in mainland and Hong Kong stocks.
Its services have been overwhelmed with demand since unveiling an artificial intelligence chatbot that it says can rival OpenAI’s ChatGPT and was developed at a fraction of the cost of competing products. Its models have since been adopted by a plethora of Chinese firms across multiple industries, even as foreign governments from Australia to the US move to block its usage over security concerns.
Last week, DeepSeek said it plans to release key code and data to the public, an unusual step to share more of its core technology than rivals such as OpenAI have done. That potentially escalates a race between the US and China to develop ever more advanced AI models.
On Wednesday, the startup said on X it was making available what it called a “DeepGEMM” library employed in the training and inference of its V3 and R1 AI models, taking one of its initial steps in that process.
