siglip2-base-patch16-224

Name: siglip2-base-patch16-224
Author: google

by google

SigLIP 2 at base size and fixed 224px, the cheap end of the family

375Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

google/siglip2-base-patch16-224

Feature URI

Overview

SigLIP 2 replaces CLIP's softmax contrastive loss with a pairwise sigmoid loss, which removes the need for a global normalization across the batch and trains stably at scale. This is the base checkpoint at fixed 224x224 input.

Against the NaFlex variants it gives up native aspect ratio handling, which matters most on documents, screenshots and wide video frames where squashing to a square distorts the spatial relationships the encoder is trying to capture. What it gains is cost: at 375M parameters it is the checkpoint you can afford to run over every frame rather than a keyframe.

Apache 2.0.

Architecture

SigLIP 2 base, 16x16 patches, fixed 224x224 input, 375,187,970 parameters. Zero-shot-image-classification pipeline with an aligned image-text embedding space.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so siglip2-base-patch16-224 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.