Gerbil
A local LLM desktop app runs GGUF models and generates images on Windows, macOS and Linux, with offline operation and CPU or GPU support.
Gerbil is a desktop app for people who want to chat with language models and generate images on their own hardware. It brings model discovery and local generation into one interface.
- Local model chat: Run GGUF language models through KoboldCpp, a backend built on llama.cpp, and import models you already have.
- Image generation: Use built-in workflow presets for Flux, Chroma, Qwen Image and Z-Image.
- Model discovery: Search Hugging Face, read model cards and download GGUF files within the app; browsing and downloads need internet access.
- Alternative chat interfaces: Launch SillyTavern for character conversations or OpenWebUI for browser-based chat, with Node.js required for SillyTavern and uv for OpenWebUI.
Gerbil is open source under AGPL-3.0 and runs on Windows, macOS and Linux, including Wayland. It works on CPU-only computers and supports GPU acceleration through CUDA, ROCm, Vulkan and Metal.
Generation stays on your machine, with no telemetry or account requirement. It can work entirely offline when the models and KoboldCpp binaries are already downloaded. The macOS app isn't verified by Apple, so the operating system can block it from opening.