Kimi-K2-Instruct-0905 is a large instruction-tuned model developed by Moonshot AI. It supports multimodal inputs including text, images, and video, along with advanced tool calling and structured generation. The model is available with open weights on Hugging Face for local deployment or use through compatible inference frameworks.
In the Multimodal & vision space, Kimi K2 Instruct 0905 takes a focused approach. It focuses on providing a capable open-source multimodal assistant with native support for images, video, and structured tool interactions. It is built as an open-source project for developers. Kimi K2 Instruct 0905 is open source under the Open Source license. It ships for the web, the command line, and API.
Behind Kimi K2 Instruct 0905 is Moonshot AI, based in China, and it first shipped in 2025. The project is developed in the open on GitHub with 11k stars. Among its 4 catalogued features are Multimodal Input, Tool Calling, and Instruction Following. It exposes integrations via a public API.
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