dinov3-vits16-pretrain-lvd1689m

Name: dinov3-vits16-pretrain-lvd1689m
Author: facebook

by facebook

DINOv3 at 21.6M parameters, small enough to run over everything

21.6Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

facebook/dinov3-vits16-pretrain-lvd1689m

Feature URI

Overview

The smallest DINOv3 checkpoint, distilled from the same 7B teacher as the larger ones. Self-supervised, no text tower, so it groups images by how they look rather than by what a caption would say about them.

Size is the whole argument here. At 21.6M parameters this runs over an entire image library on hardware that would choke on the ViT-B, which makes it the right first pass for deduplication and near-duplicate detection where you need to touch every file rather than a sample.

Use a larger checkpoint when precision on hard pairs matters more than covering the corpus.

Architecture

Vision transformer, patch size 16, DINOv3ViTModel with 21,596,544 parameters. Self-supervised training on LVD-1689M, distilled from the 7B teacher. Image-feature-extraction only: no text encoder, so it cannot answer a text query on its own.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so dinov3-vits16-pretrain-lvd1689m runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
  "https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
  {
    method: "POST",
    headers: {
      Authorization: "Bearer API_KEY",
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      collection_id: "col_your_collection",
      documents: [
        {
          document_id: "asset-00412",
          // The vector name has to match a vector index on the collection.
          vectors: { "image-embedding": yourVector },
          payload: { source_key: "archive/2026/asset-00412" },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// image_extractor@v1 runs google/siglip-base-patch16-224
// (768-d) over a bucket, with no inference of your own.