Oxide Lab
A native desktop chat application, built with Rust, Tauri, and Candle, for running large language models locally with no cloud connection or account.
Oxide Lab is a native desktop application for running large language models entirely locally. It supports several model architectures, including Llama, Qwen2, Qwen2.5, Qwen3 and its mixture-of-experts variant, Mistral, Mixtral, DeepSeek, Yi, and SmolLM2, in GGUF and SafeTensors formats, with hardware acceleration on CPU, NVIDIA CUDA, Apple Metal, Intel MKL, and Apple Accelerate.
The interface is a chat window built with Svelte 5 and Tailwind CSS, translated into English, Russian, and Brazilian Portuguese, and text generation streams as it is produced. Using it means building or downloading the application, downloading a compatible GGUF or SafeTensors model such as from Hugging Face, and loading it through the interface.
It is built with Rust and Tauri v2 on the frontend and Candle for inference, needs at least 4 GB of RAM, 8 GB or more recommended for larger models, and runs on Windows, macOS, and Linux. It is released under the Apache-2.0 license.