OCRStack adds Qwen3.5 OCR for multilingual document recognition

OCRStack started as a practical browser OCR workspace: upload a PDF or image, review the text, then download Word, TXT, or Markdown. Today we are expanding the model lineup with Qwen3.5 OCR.

You can now choose between Mistral OCR models and Qwen3.5 OCR in the same upload flow, keep one credit balance, and export in the same formats.

What shipped

• Qwen3.5 OCR is available in the OCR model picker on /tool and in PDF to Word.
• Qwen document parsing results are normalized into clean Markdown, including pipe tables instead of raw HTML table tags.
• Preview, plain-text copy, Word (.docx), TXT, and Markdown downloads all work from the same normalized result.
• Multilingual and Asian-language documents are a primary use case for the new model option.

When to choose Qwen3.5 OCR

Use Qwen3.5 OCR when your documents mix languages, include denser tables, or need stronger coverage for Chinese and other Asian scripts. Keep Mistral OCR when you already like its layout on English-heavy PDFs and want the models you have been using.

There is no single best OCR model for every file. OCRStack keeps the choice inside one workspace so you can compare outputs on your own invoices, scans, and reports.

How to try it

1. Open the OCR Tool.
2. Select Qwen3.5 OCR in the model menu.
3. Upload a PDF or image.
4. Preview the text, then download Word, TXT, or Markdown.

Open OCR Tool with Qwen3.5 OCR

What comes next

We will keep expanding model choice and task-focused pages such as invoice OCR and table extraction. If you care about a specific document type, try Qwen3.5 OCR on a real sample and tell us where the preview or export still falls short.

Release notes for this launch also live on the Changelog.

Read the changelog