Use Cases
How to get the most out of our AI extraction engines depending on your source material.
M
Cost/Benefit Model (Mistral)
Optimized for speed. Uses the `mistral-small-latest` vision capabilities to rapidly parse standard documents. Ideal for bulk workflows where cost per token matters.
- Printed Documents: PDF reports, instruction manuals, books.
- Screenshots: Chat logs, code snippets, web article captures.
- Latin Handwriting: Clear English/Spanish/French cursive or print.
- Standard Tables: Financial statements, structured CSV data from images.
G
Wide Languages Model (Gemini)
The heavy-lifter. Uses `gemini-3-flash-preview` to execute advanced spatial reasoning and multilingual text recognition. Consumes more API units but guarantees elite accuracy on hard inputs.
- Complex Non-Latin Scripts: Japanese Kanji, Mandarin, Korean Hangul, Arabic, Hindi.
- Messy/Illegible Handwriting: Doctor's notes, rapid scrawls, multi-directional writing.
- Degraded Source Material: Faded historical documents, highly compressed memes, low-light photos.
- Dense Mathematical Formulas: LaTeX extraction of scientific papers.