Gerbil
lastmod 2026-01-21created 2026-01-211 edits
A desktop app for running local LLMs
About
- A graphical UI tool designed to make it easy to run LLMs on personal hardware
- Automates complex steps such as backend management, model downloads, and hardware acceleration setup
- Runs on KoboldCpp, a fork of llama.cpp, and supports both text and image generation
- Runs natively on Windows, macOS, and Linux (including Wayland in particular)
- Works even on CPU-only systems, and supports GPU acceleration based on CUDA, ROCm, Vulkan, CLBlast, and Metal
- Built-in image generation: provides presets for major workflows such as Flux, Chroma, Qwen Image, and Z-Image
- Integrated HuggingFace search lets you browse models, view model cards, and download GGUF files directly within the app
- Can run from the terminal with the --cli option, running the same backend without the GUI and minimizing resource usage
- AGPL-3.0 license