“Kimi K3 has received far more love than we expected, and our GPUs are feeling it,” Moonshot said in a Sunday (July 19) post on social platform X. “Over the past 48 hours, demand has pushed close to the limits of our current capacity.”
By pausing new subscriptions to Kimi K3, Moonshot aims to protect the experience of its existing subscribers, who are not affected by the pause, according to the post.
Moonshot is adding capacity as quickly as it can, will reopen new subscription spots in batches, and will divide its membership plans into Kimi Membership for Kimi Web, App and Work and Kimi Code Membership for coding workflows, per the post.
“This will help us match compute more precisely and keep the experience stable,” Moonshot said in its post.
It was reported Thursday (July 16) that Kimi K3 is an open-weight model that is expected to rival Anthropic’s flagship systems while costing a fraction to run and that it could mark Moonshot’s boldest challenge yet to AI leadership in the United States.
When Moonshot introduced Kimi K3 on Thursday, the company said in a blog post that it is a 2.8 trillion-parameter model with native vision capabilities and a 1 million-token context window.
“It is the world’s first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work and reasoning,” Moonshot said in the post. “While its overall performance still trails the most powerful proprietary models, Claude Fable 5 and [OpenAI’s] GPT-5.6 Sol, Kimi K3 demonstrated frontier-level performance across our evaluation suite, consistently outperforming other tested models.”
The Associated Press reported Monday (July 20) that Kimi K3’s 2.8 trillion parameters make it the world’s largest open-source AI model. The model aims to challenge more advanced AI models, and its release impacted the stocks of U.S. tech giants whose pricing power and demand could be undercut by this and other more affordable Chinese AI models.
Reuters reported Monday that the release of Kimi K3 and other models that are increasingly capable at lower cost challenges assumptions that Chinese model developers are months behind their U.S. peers.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
