Chinese AI startup DeepSeek has reopened access to its API after halting service for nearly three weeks due to capacity constraints.

On Tuesday, the company began allowing customers to top up credits for use on its API, which lets developers build apps and services on top of cloud-hosted versions of DeepSeek’s AI. Server resources remain strained during the daytime, however, a representative for the company cautioned in a WeChat message seen by Bloomberg.

DeepSeek rose to prominence earlier this year following the release of its openly available R1 “reasoning” model, which matches or bests the performance of some of OpenAI’s top models. DeepSeek’s competitiveness has prompted OpenAI to consider open-sourcing more of its technology and “pull up” certain product releases.

As Bloomberg notes, DeepSeek’s domestic rivals are ramping up production of their models, as well. The same day DeepSeek resumed API top-ups, Chinese tech giant Alibaba launched a preview of its latest reasoning AI model, QwQ-Max, which the company plans to open-source.