Chinese AI start-up had just unveiled its most advanced artificial intelligence model yet, Kimi K3, but after that, the company has been temporarily forced to shut down its subscription operations for the same and is not accepting new subscriptions anymore. The reason for this is the fact that Kimi K3 has become too popular. The company says that its demand has surged so much that its computing infrastructure has been compromised.
Taking to X, the company posted, "Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritising compute for current members. Existing subscribed users are not affected.” The company has iterated that the older subscription already taken will not be affected by it, and the subscriptions for its membership will open up in batches now as more capacity is installed to deal with this.
Along with temporarily pausing new subscriptions, Moonshot has also announced changes to its paid plans. The company said it will now offer two separate memberships. Kimi Membership will cover Kimi Web, the mobile app and Kimi Work, while Kimi Code Membership will be designed specifically for coding-related tasks. According to Moonshot, splitting the plans will help it allocate computing resources more efficiently and ensure a more stable experience for users.
The announcement comes just days after Moonshot introduced Kimi K3 at the World Artificial Intelligence Conference in Shanghai. The company is positioning the new model as one of China's strongest rivals to leading AI models from OpenAI and Anthropic. Moonshot has also confirmed that it will release Kimi K3's open source model weights on 27 July, allowing developers to download, customise and run the model on their own systems.
What is Kimi K3?
Moonshot describes Kimi K3 as its most advanced flagship AI model so far. It is built on a 2.8 trillion-parameter Mixture of Experts (MoE) architecture and features a one-million-token context window, along with native multimodal capabilities that enable it to understand both text and images. The model is designed to handle long-running reasoning, coding and AI agent tasks.
Unlike conventional chatbots that mainly respond to questions, Kimi K3 is built to carry out complex, multi-step workflows with minimal human input. It can help write software, analyse research papers, process financial data, understand images, create presentations, and even assist with video editing.
These capabilities are powered by two new technologies called Kimi Delta Attention and Attention Residuals. Moonshot says they enable the model to process much larger volumes of information while using computing resources more efficiently.
Standing In Competition With Claude And ChatGPT
This AI model from Moonshot now stands as a serious competitor to mainstream AI models like Claude and ChatGPT from Anthropic and OpenAI, respectively. The company is of the view that its AI model performs on par with mainstream or frontier AI systems, as it was tested on several public benchmarks. It is also being claimed that Kimi K3 Max had outperformed AI models like Anthropic's Claude Opus 4.8 and OpenAI GPT-5.5 when it was tested on its own internal platform covering online research, presentation creation and financial analysis. But one must note here that results echo Moonshot's own analysis, and independent verification is still pending.
ALSO READ: Apple Music, Apple One Subscription Now More Expensive In India: Here's What You'll Pay Now
The launch comes at a time when Chinese and US tech giants are leaping towards making the most advanced AI models. Moonshot also has plans to make Kimi K3 open-source so developers can utilise and build on it for their own custom builds, while OpenAI and Anthropic keep their flagship models restricted.