llama-models

What it is and what it does

llama-models is a command-line and Python toolkit for managing Meta's Llama language models. It handles the full lifecycle: listing available models (Llama 2, 3, 3.1, 3.2, 3.3, and 4 variants), downloading weights via authenticated URLs, verifying download integrity, and removing local checkpoints. The package wraps model access from both Meta's direct distribution and Hugging Face, abstracting away authentication and URL expiration logic.

The package is primarily a download and management layer, not an inference engine itself—it prepares models for use with external frameworks like transformers or the Llama Stack. It depends on nine runtime libraries (PyYAML, Jinja2, tiktoken, Pydantic, Pillow, rich, httpx, termcolor, huggingface-hub) to handle configuration, tokenization, HTTP requests, and CLI formatting. Inference requires separate setup (PyTorch, GPUs for larger models) and is documented via example scripts in the repository.

Use it for:

• Download and cache Llama model weights locally after obtaining a Meta-signed URL, automating the authentication and expiration-handling workflow.
• List available Llama model versions and their metadata (context length, tokenizer type, acceptable-use policies) to choose the right variant for your task.
• Verify integrity of downloaded model checkpoints before loading them into inference pipelines.
• Manage multiple Llama model versions on disk, removing old checkpoints to free space.
• Retrieve prompt formatting templates for a specific model version to ensure correct chat or completion input structure.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides a CLI and Python interface to download, manage, and run Meta's Llama language models, handling authentication, model listing, verification, and local checkpoint management.

Yes, if you are building with Llama models and need a reliable CLI to download and manage checkpoints. The low install friction and absence of known vulnerabilities make it safe to add. However, the aging maintenance status (no release in ~9 months, last commit in Feb 2026) suggests the project may be in maintenance mode; verify that it still works with your target Python and dependency versions before relying on it for production automation. License treatment is unclear—confirm that the package's own license (or lack thereof) and the model licenses align with your use case.

Install

llama-models on PyPI

pip install llama-models

uv add llama-models

poetry add llama-models

Installing llama-models

Before you install

Low install friction with a pure-Python wheel and nine well-established dependencies. Maintenance status is aging—last commit was 2026-02-11 and the project has not released since 2025-11-04—so expect slower response to issues or breaking changes in upstream dependencies.

License in practice

License treatment is unclear: the package itself has no declared SPDX license, though the Llama model weights are subject to Meta's own acceptable-use policies and licenses (which vary by model version). Verify compliance with the specific model license before commercial deployment.

Quickstart

pip install llama-models
llama-model list
llama-model download --source meta --model-id <MODEL_ID>

Requires Python >=3.10 and a signed URL from Meta (obtained by accepting their license at https://llama.meta.com/llama-downloads/); links expire after 24 hours.

Verify before relying

• Whether the package's own license status affects redistribution or commercial use of downloaded models.
• Current state of GPU memory requirements for inference with different quantization modes (FP8, Int4).
• Whether aging maintenance status (last commit Feb 2026, no release since Nov 2025) signals planned deprecation or stable maturity.

Package facts

Evidence: llama_models-0.3.0-py3-none-any.whl

Tags

More Artificial Intelligence packages