AI news · October 10, 2026
OpenDocRouter puts document parsing models behind one API
OpenDocRouter accepts PDFs, images, and file URLs, returns Markdown with optional layout boxes, and starts with $5 in credits; builders can switch parsers without rewriting their document pipeline.
- new-account credits
- $5
- sync limit
- 50 pages
- async input limit
- 500 pages
LlamaIndex launched OpenDocRouter on Oct 7, and TLDR AI included it in its Oct 8 issue. The service accepts PDFs, PNGs, JPEGs, and URLs, then returns Markdown through a synchronous or asynchronous API. Each parser runs a versioned recipe, and an optional layout setting adds grounded boxes and reading order so tables, charts, forms, and headings use the same output classes.
New accounts receive $5 in credits; synchronous requests support up to 50 pages, while asynchronous requests can process up to 500 pages or 50MB. The launch lists open parsers such as MinerU2.5-Pro alongside Claude, Gemini, GPT, and other models. That is more flexible than wiring each model yourself, but the new service still needs a quality and cost comparison with LlamaParse, direct OCR, and your own document set.
What you can do with it
Create an account and try the curl example on a few invoices or reports; new accounts receive $5 in credits. Start with one parser and layout off, then compare Markdown, tables, and total cost with your current OCR before routing production files.
Our take
This is useful plumbing for a small team because parser choice becomes a setting instead of a rewrite. The catch is that one API does not remove the need to inspect tables, charts, and sensitive documents, and the launch's benchmark and pricing are still new.
Links Try the API
Source: LlamaIndex — Made With Models writes the brief; the reporting is theirs.