Practical, runnable recipes for building with Eximius Labs open models. Each cookbook is a small, self-contained project you can run and adapt.
- kimi-k3-multimodal-memory — give a Kimi K3 agent a multimodal memory it can search across text, images, and audio, using fusion-embedding-2. K3 reasons over what it retrieves, including sounds it cannot hear directly. Works with any OpenAI-compatible agent model; K3 is the timely example.
More cookbooks for other models are on the way.
Eximius Labs builds open-weight multimodal models that run on your own hardware: the memory and perception layer for physical AI. One shared embedding space spans text, image, video, audio, thermal (the Ember pack), and motion and IMU (the Tremor pack), alongside a perception layer (Fusion Perception).
- Models: https://huggingface.co/EximiusLabs
- Code: https://github.com/Eximius-Labs
- Paper: https://arxiv.org/abs/2607.18666
Apache-2.0 (cookbook code). Model weights each cookbook downloads are released separately under their own licenses.