Moonshot AI has released Kimi K3, an open-weight AI model with 2.7 trillion parameters. The company says it's the largest open-weight model ever made public.
What open-weight means
Open-weight models release the trained parameters for anyone to download and use. That's different from closed models, which keep the weights private and only offer access through an API. Researchers and developers can run Kimi K3 on their own servers, fine-tune it for specific tasks, or study its internal behavior.
Scale of the model
2.7 trillion parameters is a massive size. No other open-weight model has reached this scale. The parameter count is a key metric for AI model capacity. Larger models can handle more complex tasks, but they also require significant compute resources to run. The model will likely be used by organizations with access to large clusters of GPUs.
Moonshot AI has not shared benchmark results or details on the training data for Kimi K3. The announcement focused on the model's size and open-weight nature. The model is available for download from the company's website. Researchers can now access it and start experimenting with its capabilities.
