GPT vs Marker
Pick GPT when you need to read handwriting, pull data into custom JSON, or answer questions about a document, and you accept per-token billing through a hosted API. Pick Marker when you want to convert PDFs to Markdown on your own hardware, especially books and papers with equations, and your company falls inside the weight license terms.
GPTopenai/gpt | Markerdatalab/marker | |
|---|---|---|
| Type | Vision model | OCR API |
| Price | 11 credits / page | 14 credits / page |
| Per 1,000 pages | $11.00 | $14.00 |
| Typical speed | 2.4 s / page | 2.5 s / page |
| Best for | Messy scans, custom instructions | Books and research papers |
| Inputs | PDF, images | PDF, images, DOCX |
| Languages | 50+ languages | 90+ languages |
| Max pages per request | 100 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | ✓ Yes |
| 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 GPT if
- It handles handwritten notes: one HN user reported about 5% errors on messy cursive, versus about 30% for Mistral OCR.
- It can extract into custom JSON and answer questions about a document, because you control the prompt.
- PDF input sends both extracted text and page images to the model, and files up to 50 MB per request are accepted.
Choose Marker if
- You can run it locally on GPU, CPU or Apple Silicon, so no page leaves your machine.
- It is built for books and papers with equations, and it can output LaTeX, HTML, JSON block trees and RAG chunks as well as Markdown.
- Cost is predictable: free self-hosted (you pay for compute), or about 5,000 free pages per month on the hosted API.
Key differences
| GPT | Marker | |
|---|---|---|
| How you run it | Hosted OpenAI API only | Self-hosted on GPU, CPU or MPS, or Datalab's hosted API |
| Licence | Proprietary | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M license, free for research, personal use and startups under $5M |
| Pricing | Token-based per model; page images add input tokens, so cost per page is not fixed | Free self-hosted within the weight license; hosted API has about 5,000 free pages per month |
| Output formats | Markdown, JSON, plain text | Markdown, HTML, JSON, chunks, LaTeX, images |
| Benchmark | GPT-5.2 scored 85.50 on OmniDocBench v1.5 and 86.59 on v1.6, both independent | 83.7 on olmOCR-bench and 85.9 on OmniParseBench in vllm mode, both reported by Datalab itself |
| Coordinates and confidence | No bounding boxes or confidence scores in the standard response | Not published |
| Known limit | Not suited to cheapest bulk OCR or pipelines that need word coordinates | Weight license blocks free commercial use for companies above $5M; CPU fast mode drops to 66.6 on olmOCR-bench |
Sources
- OpenAI file inputs docs
- OmniDocBench leaderboard
- HN Mistral OCR thread
- OpenAI API pricing
- Marker GitHub README
- Datalab pricing
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker 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": "openai/gpt","fallbacks": ["datalab/marker"],"file_url": "https://example.com/document.pdf"}