LibreDINOv3vitl16

A mirror of Meta's DINOv3 ViT-L/16 image encoder, facebook/dinov3-vitl16-pretrain-lvd1689m (pretrained on LVD-1689M), for LibreYOLO.

LibreYOLO's TRELLIS.2 image-to-3D model reads its input through this encoder, as Microsoft's TRELLIS.2 does. The mirror lets LibreGen("trellis2") run without a Hugging Face login or an access request.

This is a mirror, not a new model

The weights are Meta's. Nothing was retrained, distilled, fine-tuned or cast: config.json, preprocessor_config.json and model.safetensors are byte-identical to upstream (float32, the precision TRELLIS.2 runs the encoder in).

• Upstream repository: facebook/dinov3-vitl16-pretrain-lvd1689m
• Upstream revision: ea8dc2863c51be0a264bab82070e3e8836b02d51
• SHA-256 of every file is listed in NOTICE.

License: read this before use

These weights are not covered by LibreYOLO's MIT license. They are redistributed under Meta's DINOv3 License, included here as LICENSE. The license permits redistribution on the condition that it is passed through: the weights may only be distributed under the same agreement, with a copy of it. Using or distributing them binds you to it.

Restrictions that pass through, summarized but not replacing the text:

• Use must comply with applicable laws, including Trade Controls and privacy and data protection law, and you must not be the target of Trade Controls.
• No activities subject to ITAR or end uses prohibited by Trade Controls, including military or warfare purposes, nuclear industries or applications, espionage, or the development or use of guns or illegal weapons.
• No reverse engineering or decompiling of the materials.
• Research publications that use these weights must acknowledge them.
• Meta may modify the agreement, with changes effective immediately.

Meta gates its own repository behind an access form. This mirror is not gated; users who need Meta's own access process should obtain the weights from the upstream repository.

Usage

With LibreYOLO, which downloads it automatically:

from libreyolo import LibreGen

LibreGen("trellis2")("chair.jpg").save("chair.obj")

With transformers:

from transformers import AutoModel

model = AutoModel.from_pretrained("LibreYOLO/LibreDINOv3vitl16")

Citation

@misc{simeoni2025dinov3,
  title={{DINOv3}},
  author={Sim{\'e}oni, Oriane and Vo, Huy V. and Seitzer, Maximilian and Baldassarre, Federico and Oquab, Maxime and Jose, Cijo and Khalidov, Vasil and Szafraniec, Marc and Yi, Seungeun and Ramamonjisoa, Micha{\"e}l and Massa, Francisco and Haziza, Daniel and Wehrstedt, Luca and Wang, Jianyuan and Darcet, Timoth{\'e}e and Moutakanni, Th{\'e}o and Sentana, Leonel and Roberts, Claire and Vedaldi, Andrea and Tolan, Jamie and Brandt, John and Couprie, Camille and Mairal, Julien and J{\'e}gou, Herv{\'e} and Labatut, Patrick and Bojanowski, Piotr},
  year={2025},
  eprint={2508.10104},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2508.10104},
}