OCRmyPDF vs InternVL3.5
Pick OCRmyPDF to make scanned PDF archives searchable and PDF/A compliant on CPU-only machines. Pick InternVL3.5 when you need Markdown or JSON out of documents, such as document question answering, and you have GPUs to run it. OCRmyPDF keeps the original PDF and adds a text layer, while InternVL3.5 reads page images and returns new text.
OCRmyPDFocrmypdf/ocrmypdf | InternVL3.5opengvlab/internvl | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 1.0 s / page | 3.0 s / page |
| Best for | Searchable PDFs from scans | Self-hosted document understanding |
| Inputs | PDF, images | images |
| Languages | 100+ languages (via Tesseract) | Multilingual |
| Max pages per request | 1,000 | 500 |
| Tables | No | ✓ Yes |
| Handwriting | No | No |
| Math | No | ✓ Yes |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose OCRmyPDF if
- You need validated searchable PDF/A output for archiving, and the OCR layer is inserted losslessly where possible.
- You only have CPUs, since it spreads work across all CPU cores and needs no GPU.
- You need 100+ languages through Tesseract language packs, in an office scanner pipeline.
Choose InternVL3.5 if
- You want Markdown or JSON output, or document question answering, which OCRmyPDF does not produce.
- You want one model family in nine sizes from 1B to 241B, so you can trade accuracy for cost.
- You have GPUs available: models up to 30B fit on a single A100, and the MoE variants 20B-A4B and 30B-A3B activate few parameters per token.
Key differences
| OCRmyPDF | InternVL3.5 | |
|---|---|---|
| What it is | Command-line tool that adds a Tesseract OCR text layer to scanned PDFs | Open-weight vision language model family from OpenGVLab, 1B to 241B-A28B parameters |
| Licence | MPL-2.0 | Apache-2.0, built on Apache-2.0 Qwen3 |
| Hardware | CPU only; distributes work across all CPU cores | Up to 30B runs on a single A100; 38B needs two A100s; 241B-A28B needs eight A100s |
| Output formats | Searchable PDF (PDF/A by default) and plain text | Plain text, Markdown and JSON |
| Layout and tables | Adds no layout, table or Markdown output | No bounding boxes or layout output by default |
| Benchmarks | Not published | OCR scores appear only as chart images on the model card, not as text values |
| Known failure | Hangs on PDFs whose forms share one Resources dictionary; --skip-text with default PDF/A output removes form fields | 38B needs two A100s with tensor parallelism; the largest model needs eight A100s |
Sources
- OCRmyPDF GitHub README
- OCRmyPDF docs
- Issue #1790 form fields
- Issue #1763 hang
- Hugging Face model card (8B)
- InternVL GitHub
What each one is
OCRmyPDF
Command-line tool that adds a searchable text layer to scanned PDFs, using Tesseract.
OCRmyPDF detailsInternVL3.5
Shanghai AI Lab's open vision-language family (1B to 241B, August 2025), strong on OCR and document benchmarks.
InternVL3.5 detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "ocrmypdf/ocrmypdf","fallbacks": ["opengvlab/internvl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
OCRmyPDF alternativesInternVL3.5 alternativesBest OCR models in 2026