# Inference Snaps

A collection of Ubuntu snap packages that install a locally optimized runtime and model weights for a chosen open language model in one command, tuned automatically for the host hardware.

Each inference snap bundles a hardware-aware runtime with its model weights, so installing one snap sets up a working local AI endpoint without separately managing dependencies or driver-specific builds. Snaps currently cover models including DeepSeek R1, several Gemma and Qwen variants, GLM 4.7 Flash, GLM-OCR, several Nemotron variants, Nomic Embed Text, Phi 4 and SmolLM2, with capabilities noted per model such as text, thinking, vision and tool use.

After installing an accelerator driver and a chosen model snap, a status command reports readiness and a chat command starts an interactive session; each snap exposes a local, standard API for other software to call. Snaps run entirely on the local machine once installed, with consistent, isolated environments across supported Linux distributions.

Website: https://documentation.ubuntu.com/inference-snaps