docTR vs Pixtral
Pick docTR when you need word-level text with coordinates from invoices and receipts and you want a small, trainable OCR library you run yourself in Python. Pick Pixtral only if you already use it through the Mistral API or want question answering over charts and documents. The main split is that docTR is a purpose-built OCR pipeline, while Pixtral is a deprecated general vision language model that can miss text.
docTRmindee/doctr | Pixtralmistral/pixtral | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 0.9 s / page | 3.0 s / page |
| Best for | Word-level OCR with positions | Questions about document images |
| Inputs | PDF, images | images |
| Languages | Mainly Latin scripts | Multilingual |
| Max pages per request | 500 | 500 |
| Tables | No | No |
| Handwriting | No | No |
| Math | No | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | 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 docTR if
- You need JSON output with page, block, line and word levels and coordinates, or bounding boxes, for custom extraction.
- You want to choose or train your own detection and recognition architectures, with straight or rotated boxes, in Python on PyTorch.
- You want a ready Docker image, Apache-2.0 licence and a key information extraction predictor, with no per-token API cost.
Choose Pixtral if
- You answer questions over charts and documents, where Mistral reports 90.7 ANLS on DocVQA and 81.8 on ChartQA.
- You already have an integration on the Mistral API, where the launch price was $0.15 per million input and output tokens.
- You need many images in one request, which the 128K token context allows, and you want Markdown or JSON output from a prompt.
Key differences
| docTR | Pixtral | |
|---|---|---|
| What it is | OCR library with two-stage text detection and recognition | General vision language model, 12B decoder plus 400M vision encoder |
| Licence | Apache-2.0 | Apache-2.0 open weights |
| How you run it | Local Python on PyTorch, with Docker images; CPU or NVIDIA GPU, and GPU images need CUDA 12.2 or newer on the host | Mistral API as pixtral-12b-2409, or self-hosted open weights |
| Price | Free, self-hosted (you pay for compute) | $0.15 per million input and output tokens at API launch; free to self-host (you pay for compute) |
| Output formats | JSON, plain text, bounding boxes; no table, formula or Markdown output | Plain text, Markdown, JSON |
| Benchmarks | Not published | DocVQA 90.7 ANLS, ChartQA 81.8, MMMU 52.5, all reported by Mistral |
| Known failure | Split-crop merging misjoins readings across repeated characters; cropping images gives poorer results; ONNX export of MASTER truncates decoding steps | Does not always get OCR near correct; accuracy drops reported on handwriting, small fonts and non-Latin scripts; Q4 quantization degrades small text; Mistral marks it deprecated and unmaintained |
Sources
- docTR GitHub README
- docTR documentation
- Issue #2229
- Mistral announcement
- Mistral Sept 2024 pricing
- HN launch thread
What each one is
docTR
Open-source OCR library from Mindee with separate text detection and recognition models.
docTR detailsPixtral
Mistral's open-weight vision models, Pixtral 12B and Pixtral Large. Mistral's API no longer serves them, so they run self-hosted.
Pixtral 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": "mindee/doctr","fallbacks": ["mistral/pixtral"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
docTR alternativesPixtral alternativesBest OCR models in 2026