PaddleOCR-VL-1.5

Name: PaddleOCR-VL-1.5
Author: PaddlePaddle

by PaddlePaddle

Vision-language OCR handling text, tables, formulas, charts, and 109 languages in 0.9B params

15Kdl/month

661likes

959Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

PaddlePaddle/PaddleOCR-VL-1.5

Feature URI

mixpeek://image_extractor@v1/paddle_ocr_vl_15_v1

Overview

PaddleOCR-VL 1.5 replaces the traditional OCR pipeline (detect → recognize → layout) with a single vision-language model that understands document structure natively. At 0.9B parameters, it handles text recognition, table extraction, formula parsing, chart understanding, seal detection, text spotting, and 109 languages, including rare scripts like Tibetan and Bengali.

On Mixpeek, PaddleOCR-VL replaces brittle multi-stage OCR pipelines with a single model call that produces structured output from any document type. Its robustness to scanning artifacts, skew, and poor lighting makes it reliable for real-world document ingestion.

Architecture

Vision-language model with document-specific pretraining. 0.9B parameters. Unified multi-task architecture handles text detection, recognition, layout analysis, table extraction, and formula parsing in a single forward pass. Robust to image degradation (scanning, warping, screen capture).

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so PaddleOCR-VL-1.5 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 model produces text, so it lands in payload. Give the
          // collection a text vector index and embed that text to make it
          // searchable rather than only filterable.
          payload: { extracted_text: modelOutput, source_key: "archive/2026/asset-00412" },
          vectors: { "text-embedding": embeddingOfModelOutput },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// universal_extractor@v1 runs google/gemini-embedding-2
// (3072-d) over a bucket, with no inference of your own.