Kimi-K3 is a large multimodal model from Moonshot AI released on Hugging Face under a permissive license. It supports image-text-to-text tasks and can be used with popular libraries such as Transformers and vLLM. The model is provided in Safetensors format with custom code support, enabling developers to run it locally, in notebooks, or via self-hosted inference servers. It is intended for researchers and engineers building applications that require vision-language capabilities on their own infrastructure.
In the Multimodal & vision space, Kimi K3 takes a focused approach. It focuses on running advanced multimodal AI inference locally or in self-hosted environments without relying on proprietary cloud APIs. Kimi K3 is an open-source project aimed at developers and AI researchers. Kimi K3 is open source under the Open Source license. It ships for the web, the command line, and API, and it can be self-hosted.
Moonshot AI builds and maintains Kimi K3. Among its 4 catalogued features are Transformers Integration, vLLM Support, and Safetensors Format.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do