Kimi releases K3.1 Pro, a 2.8-trillion-parameter multimodal model

Kimi has released K3.1 Pro, a large multimodal AI model with 2.8 trillion parameters that processes images and text as input and generates text as output. The model weights are now publicly available on Hugging Face, allowing researchers and developers to download, explore and run the model themselves.

This release follows the trend of major AI labs and companies open-sourcing or releasing significant model weights, making frontier-scale models more accessible to the broader AI community. The scale and capabilities of K3.1 Pro position it as a notable entry in the emerging landscape of large multimodal models.