olmOCR vs Nougat
Pick olmOCR for large-scale PDF to Markdown in a commercial product, especially with mixed printed and handwritten pages or a RAG pipeline. Pick Nougat only for non-commercial work on English arXiv-style papers where you want LaTeX output. The main split is licensing and scope: olmOCR is Apache-2.0 and covers more document types, while Nougat's weights are CC-BY-NC and it is built for scientific papers.
olmOCRallenai/olmocr | Nougatmeta/nougat | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 1.5 s / page | 3.0 s / page |
| Best for | Bulk PDF to text for datasets | arXiv-style papers |
| Inputs | PDF, images | PDF, images |
| Languages | Mainly English | Mainly English |
| Max pages per request | 1,000 | 200 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | No |
| 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 olmOCR if
- Its toolkit is Apache-2.0, and weights, training data and code are all released, so you can use it commercially.
- It handles equations, tables and handwriting, and strips headers and footers in reading order, so it suits mixed pages beyond papers.
- Self-hosted cost is under $200 per million pages per Ai2, or you can call a hosted endpoint such as DeepInfra at $0.09 input / $0.19 output per million tokens.
Choose Nougat if
- It is trained on arXiv and PubMed Central documents, so it fits English scientific papers.
- It outputs LaTeX for equations and tables in Mathpix-compatible .mmd Markdown, which suits math-heavy PDFs.
- It ships small and base checkpoints with a simple CLI and API, and runs as a plain Python package.
Key differences
| olmOCR | Nougat | |
|---|---|---|
| Licence | Apache-2.0 (toolkit) | Code MIT; weights CC-BY-NC, no commercial use |
| How you run it | Self-host with vLLM on a GPU (12 GB VRAM or more), or use third-party hosted endpoints | Self-hosted Python package only |
| Price | Under $200 per million pages self-hosted; DeepInfra $0.09 / $0.19 per million tokens | Free for non-commercial use; you pay for compute |
| Output formats | Markdown, JSON | Markdown (Mathpix-style .mmd), LaTeX |
| Benchmark | 82.4 on olmOCR-Bench (v0.4.0), above Marker (76.1), MinerU 2.5.4 (75.2) and Mistral OCR (72.0) | Not published |
| Languages and handwriting | Handles handwriting and mixed printed pages | Works best on English; Chinese, Russian and Japanese do not work |
| Known failure | Out of memory on a 12 GB GPU (issue #424); can get stuck on the last few pages (issue #433); image bounding boxes are unreliable (issue #437) | Repetition loops on 1.5% of test pages in the paper, more on out-of-domain documents; looping pages can be dropped as MISSING_PAGE (issue #11) |
Sources
What each one is
olmOCR
Open-source vision model from Ai2 (often searched as olm ocr) that turns PDFs into clean, reading-order text. Latest release: olmOCR 2.
olmOCR 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": "allenai/olmocr","fallbacks": ["meta/nougat"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
olmOCR alternativesNougat alternativesBest OCR models in 2026