Download and Manage Models

This guide explains how to find models available on the current hardware and how to download, inspect, and remove local models.

Model Names and Inference Platforms

Model Names and Types

LlamaPi accepts model names with or without a parameter size:

Models currently have two types:

Currently Supported Inference Platforms

One model may have variants for different inference platforms and hardware. LlamaPi normally selects a variant for the current hardware. To select one manually, use --platform <platform>/<chip>.

LlamaPi detects the current hardware and available inference platforms. Future releases will add support for more hardware and inference platforms.

Find and Download Models

Find Models Available on the Current Hardware

llamapi list --online

This command lists all downloadable models available on the current hardware.

Download a Model

Run the following command to automatically select the recommended platform and download the corresponding model:

llamapi pull qwen3:4b

You can add the --platform option to download the model for a specified inference platform:

llamapi pull qwen3:4b --platform rkllm/rk3588

After the download, LlamaPi validates the model files and displays the model size.

Manage Local Models

List Local Models

llamapi list

Local models are stored in this directory by default:

/var/lib/llamapi/models

Remove a Local Model

llamapi rm qwen3:4b

Remove a model variant for a specific inference platform:

llamapi rm qwen3:4b --platform rkllm/rk3588

If the model is running, LlamaPi asks whether to unload it first. After removing the local files, download the model again before using it.

Continue with Run and Deploy Models. See the Terminal Command Guide for complete command options.