MINICPM-OCR 1.0

Structured OCR, ready for your workflow

Review documents in the browser, export structured results, or explore the official MiniCPM-OCR tools for your own deployment.

Recognize in the browser

Upload an image or PDF in our workspace. Recognition runs through a private server integration with ModelBest; no provider API key is exposed to your browser.

Document parsing

Explore the open-source model

For a custom deployment, start with the official OpenBMB repository and model card. Follow their requirements, licensing and installation instructions.

pip install "minicpm-ocr[all]"

GitHub Read the official model card

Coordinates stay with the content

JSON exports include page dimensions, reading order, region types and bounding boxes. Coordinates are measured in pixels on the upright page; x1, y1, x2, y2 identify each region.

{
  "pages": [
    {
      "index": 0,
      "width": 1200,
      "height": 1600,
      "blocks": [
        {
          "id": "page-0-block-0",
          "order": 1,
          "label": "doc_title",
          "bbox": [
            100,
            80,
            800,
            150
          ],
          "regions": [
            [
              100,
              80,
              800,
              150
            ]
          ],
          "content": "Document title",
          "status": "ok"
        }
      ]
    }
  ]
}

About API access

This service currently offers the browser workspace and result exports. A public developer API is not included in these plans. The official ModelBest API is a separate service with its own account and billing.