Open-source inference runtime
Northloom product
Krasis
Run larger models on the hardware you already understand.
Krasis lets system memory work alongside Nvidia VRAM, so a model does not have to fit entirely on the GPU. It gives smaller teams a practical way to serve much larger models without moving straight to multi-GPU infrastructure.
- Free and open source
- Native Linux and Windows support
- Nvidia hardware
- Public repository with hundreds of stars
