(Jan 21): China’s Zhipu is limiting access to its coding assistant after strong demand for a new AI model syphoned off computing resources.
The company will on Friday (Jan 23) start taking only 20% of its current daily new subscriptions to its GLM Coding Plan, it said in a WeChat statement. Existing automatic renewals of the plan, which is similar to Anthropic PBC’s Claude, will not be affected, it added.
Zhipu said its recent launch of its GLM-4.7 large language model triggered a surge of users, resulting in temporary computing resource constraints.
The Beijing-based developer, which staged a high-profile initial public offering in Hong Kong this month, is among a group of promising Chinese contenders squaring off with Anthropic and OpenAI. US sanctions have constrained its access to top-tier Nvidia Corp AI accelerators, pushing the company to work more closely with domestic provider Huawei Technologies Co.
This month Zhipu also released a new multimodal model that it says is the country’s first to be fully trained using domestic chips. The company said that its open-source image generation model, GLM-Image, is the first state-of-the-art multimodal model to complete training using Huawei’s Ascend chips.
It’s common for nascent online service providers to encounter demand spikes. After becoming a global sensation last year, Chinese startup DeepSeek was forced to restrict access to its application programming interface service to manage its server capacity.
See also: China to set lower growth target as economy weakens, SCMP reports
Uploaded by Arion Yeow
