TrOCR vs InternVL3.5
Pick InternVL3.5 for full pages and PDFs, because TrOCR only reads single text-line images and has no page or layout output. Pick TrOCR only when you already have cropped lines, such as English handwriting or receipt lines, or want to fine-tune a recognizer on your own line images.
TrOCRmicrosoft/trocr | 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 | 1.2 s / page | 3.0 s / page |
| Best for | Handwritten text lines | Self-hosted document understanding |
| Inputs | images | images |
| Languages | English | Multilingual |
| Max pages per request | 100 | 500 |
| Tables | No | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | No | ✓ Yes |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | ✓ 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 TrOCR if
- The large model reports 2.89 character error rate on the IAM handwriting set, per the paper.
- It has separate checkpoints fine-tuned for handwritten (IAM), printed (SROIE) and scene text.
- It loads in a few lines with Hugging Face transformers and is MIT licensed, which suits fine-tuning on custom line images.
Choose InternVL3.5 if
- It reads whole pages and answers questions about documents, and it can return plain text, Markdown or JSON.
- It comes in nine sizes from 1B to 241B, so you can trade accuracy for cost, and models up to 30B fit on a single A100.
- It uses the Apache-2.0 license and can run through Transformers, LMDeploy or vLLM.
Key differences
| TrOCR | InternVL3.5 | |
|---|---|---|
| Input it handles | Single text-line images only; you need a separate line detector | General vision language model that takes document images and questions |
| Output formats | Plain text | Plain text, Markdown, JSON |
| Layout and tables | No layout, table or page-level output | No bounding boxes or layout output by default |
| Languages | Released checkpoints are English-focused; other languages need fine-tuning | Not published |
| Model size | Base 334M, large 558M parameters | 1B to 241B-A28B; the 8B variant is 8.5B |
| Licence | MIT (microsoft/unilm repository) | Apache-2.0 |
| Published OCR benchmark | IAM handwriting CER 2.89 (TrOCR-large), reported by the authors | Not published as text; the model card shows OCR scores only as chart images |
Sources
- TrOCR paper (arXiv 2109.10282)
- trocr-large-handwritten model card
- Transformers TrOCR docs
- Hugging Face model card (8B)
- InternVL GitHub
What each one is
TrOCR
Transformer text recognition model. Reads single text lines, including handwriting.
TrOCR 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.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "microsoft/trocr","fallbacks": ["opengvlab/internvl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
TrOCR vs MiniCPM-V 4.5TrOCR vs PixtralolmOCR vs InternVL3.5Kraken vs InternVL3.5TrOCR vs MonkeyOCRPaddleOCR-VL vs TrOCRTrOCR vs SmolDoclingEasyOCR vs TrOCR
TrOCR alternativesInternVL3.5 alternativesBest OCR models in 2026