Moondream

Moondream detects objects, returns points and answers image questions.

Tasks: detection, point detection
Sizes: 2, 3 at 384 to 512 px
Install: pip install "libreyolo[vlm]"
Support tier: Sibling tier, since v1.6.0. A separate product surface with its own factory and contract.
Licenses: Code MIT, weights Apache-2.0 (2); BSL 1.1 (3). Commercial use

Install

bash

pip install "libreyolo[vlm]"

Predict

Python

from libreyolo import LibreVLM, SAMPLE_IMAGE model = LibreVLM("moondream-2", device="cpu")model.set_classes(["person", "building"])result = model(SAMPLE_IMAGE)print(result.boxes)

The default selects Moondream 2. moondream-3 names the BSL 1.1 Moondream 3 model, but it does not load in 1.6.0: the LibreYOLO/LibreMoondream3 mirror is missing weight shards. Use task="point" for points and chat() for native text responses. Training is not supported.

Licensing

Check the license on the Hugging Face repository of the specific weights you download. Every checkpoint in the LibreYOLO org carries one, and they are not always the same across a family. That repository is the authoritative source; the summary below describes what applied when this page was last verified.

This is a description of the licenses involved, not legal advice. If the answer matters commercially, read the licenses yourself and take your own counsel.

Original work: Moondream, Moondream
Upstream license: Apache-2.0 (2); BSL 1.1 (3)
Upstream source: huggingface.co/moondream
LibreYOLO code: MIT
Weights: Apache-2.0 (2); BSL 1.1 (3), republished at huggingface.co/LibreYOLO
Interpretation: Moondream 2 uses Apache-2.0. Moondream 3 uses BSL 1.1 with the publisher's additional-use conditions; it is not an unrestricted Apache-2.0 checkpoint.