OCRRouter

GPT vs Chunkr

Pick GPT when you need handwriting, custom JSON extraction or questions answered over a document, and you can accept a hosted API with no word coordinates. Pick Chunkr when you need layout-aware parsing with bounding boxes and chunks for RAG, or want to self-host. The main split is a general language model that you prompt versus a parsing service that returns structured layout.

GPTopenai/gptChunkrchunkr/chunkr
TypeVision modelOCR API
Price11 credits / page14 credits / page
Per 1,000 pages$11.00$14.00
Typical speed2.4 s / page3.0 s / page
Best forMessy scans, custom instructionsLayout-aware chunks for RAG
InputsPDF, imagesPDF, images, DOCX
Languages50+ languagesMultilingual
Max pages per request1001,000
Tables✓ Yes✓ Yes
Handwriting✓ YesNo
Math✓ Yes✓ Yes
Bounding boxesNo✓ Yes
ConfidenceNoNo
FormsNoNo
Multilingual✓ Yes✓ Yes
On OCRRouterAvailableAvailable

Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.

Which to choose

Choose GPT if

  • An HN user reported about 5% errors on messy cursive versus about 30% for Mistral OCR, so it suits handwritten notes.
  • You can prompt it to extract data into custom JSON and to answer questions about the document.
  • PDF input sends both extracted text and page images to the model, and files up to 50 MB are accepted per request.

Choose Chunkr if

  • It returns OCR with bounding boxes, which GPT's standard response does not.
  • You can self-host it with Docker (GPU, CPU-only and Mac ARM options) under AGPL-3.0, or use the cloud API.
  • It is built for semantic chunking of PDFs and slides, and it accepts PDF, PPT, Word and images.

Key differences

GPTChunkr
DeploymentHosted API onlyCloud API or self-hosted via Docker
LicenceProprietaryAGPL-3.0 or commercial licence; closed-source products need the commercial one
PriceToken-based per model; page images add input tokens, so cost per page is not fixedHosted credit-based plans (official prices not verified); open-source version is free but you pay for compute and an LLM endpoint
Output formatsMarkdown, JSON, plain textHTML, Markdown, JSON, bounding boxes
Word coordinatesNo bounding boxes or confidence scores in the standard responseOCR with bounding boxes
BenchmarkGPT-5.2 scored 85.50 on OmniDocBench v1.5 and 86.59 on v1.6, versus 90.33 for Gemini 3 Pro on v1.5Not published
Known limitsCost varies with page images; no word coordinatesOpen-source version uses less accurate community models, has no Excel support, and still needs an external LLM endpoint

Sources

What each one is

GPT

General vision model. Good at messy scans and following extraction instructions.

GPT details

Chunkr

Document parsing API that segments and chunks pages for RAG. Self-hostable version available.

Chunkr details

Try 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": "openai/gpt","fallbacks": ["chunkr/chunkr"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

GPT alternativesChunkr alternativesBest OCR models in 2026