janas-get

Download and convert Janas-LLM catalog models

TLDR

List the models the catalog knows

$ janas-get list

Download a converted model into the default folder

$ janas-get [gemma-4-e4b]

Put files in a different directory

$ janas-get [qwen3-4b] --dir [./models]

Download the original GGUF and convert it locally

$ janas-get [qwen3-4b] --from-gguf

Keep the GGUF after conversion

$ janas-get [qwen3-4b] --from-gguf --keep

Fetch a Hugging Face file by path

$ janas-get [hf:owner/repo/file.jns]

Skip the model's prediction file or draft

$ janas-get [qwen3-4b] --no-mtp

SYNOPSIS

janas-get listjanas-get model [--dir DIR] [--no-mtp] [--from-gguf [--keep]]janas-get hf:owner/repo/file [--dir DIR] [--keep]

janas-get downloads a model already converted to Janas's .jns format from the project's Hugging Face account ([prabanta-dev](https://huggingface.co/prabanta-dev)), with its prediction file or draft when the catalog names one, and checks the SHA-256 of every file against MODELS.md in the Janas tree.Files go to `~/.local/share/janas/models`, or wherever JANAS_MODELS or --dir says. A download stopped halfway resumes; a file already there is checked rather than fetched again. janas-chat then finds the model by catalog name.By default the converted file is fetched. --from-gguf takes the original GGUF the catalog names, downloads it, and converts it on this machine to the same fingerprints; the GGUF is deleted once converted unless --keep. janas-get uses that path by itself when a converted file is not on Hugging Face.hf:owner/repo/file fetches any `.jns` or `.gguf` from Hugging Face, checked against the SHA-256 Hugging Face lists for it.`janas-get list` prints each catalog name, download size, and the smallest machine that model is meant for.

PARAMETERS

list

Print the catalog.

model

A catalog name such as qwen3-4b or gemma-4-e4b.

hf:owner/repo/file

A Hugging Face object (`.jns` or `.gguf`).

--dir DIR

Where the models go (default JANAS_MODELS, or `~/.local/share/janas/models`).

--no-mtp

Without the model's prediction file or draft.

--from-gguf

Download the original GGUF and convert it here.

--keep

Keep the downloaded GGUF once converted.

CAVEATS

Converted models are large (from under 1 GB to about 48 GB for Qwen3-Next-80B). Conversion needs room for both the GGUF and the `.jns` at once, about twice the model. Keep `.jns` files on an NVMe SSD: experts are read from disk while the model answers. A converted `.jns` is a derivative of the model, not of Janas, and keeps the licence its authors gave it. IQ2 and IQ1 GGUF types are not converted.

HISTORY

janas-get ships with Janas, written by Maurizio "camauri" Cammalleri (first public release September 2026, GNU GPL 3 or later).

RESOURCES

Source code · Documentation