Chinese artificial intelligence company Moonshot AI has released the model weights for its recently launched Kimi K3 large language model, allowing developers and enterprises to deploy and customize the model for their own applications under an open-weight licensing framework.
The release follows the timeline previously announced by the company and comes amid an ongoing global debate over the role of open-weight AI models versus proprietary systems developed by major U.S. technology companies.
Model Weights Now Publicly Available
The Kimi K3 model has been released through the AI development platform Hugging Face, with both the model weights and source code made available under the Kimi K3 License.
The license permits users to access, modify and build upon the software, including the model weights, configuration files, inference and training code, without charge.
However, Moonshot AI has included commercial licensing provisions. Organisations operating the model as a service business and generating more than $20 million in revenue over any consecutive 12-month period will be required to enter into a separate commercial agreement with the company.
Large-Scale AI Model
Moonshot AI has described Kimi K3 as the first open-weight AI model featuring 2.8 trillion parameters, making it one of the largest publicly available models of its kind.
The model also supports a 1 million-token context window, enabling it to process and analyse significantly larger documents and conversations than many existing AI systems.
In addition to the model weights, Moonshot AI has released several components of its AI infrastructure, including high-performance attention kernels, a Mixture-of-Experts (MoE) communication library and infrastructure designed to support large-scale AI agents.
Growing Debate Over Open AI Models
The release has intensified discussions within the AI industry over whether open-weight models developed by Chinese companies can compete with closed, proprietary systems offered by U.S.-based AI developers.
Open-weight models allow developers greater flexibility to deploy and fine-tune AI systems locally while reducing dependence on proprietary cloud-based services.
The launch also comes as several major U.S. technology companies continue to debate the balance between open AI development and commercial proprietary models.
Competitive Performance
According to Moonshot AI, Kimi K3 delivers frontier-level performance across multiple benchmark evaluations.
While the company acknowledged that the model currently trails leading proprietary systems such as Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 Sol on overall performance, it said Kimi K3 offers a strong combination of capability, openness and scalability for developers and enterprises.
The release further reflects the rapidly evolving global AI landscape, where open-weight models are emerging as an increasingly important alternative for research, enterprise deployment and AI innovation.
