Cohere · Stable
Cohere Parse 5
Cohere's compact document parser for turning enterprise PDFs, presentations, and scans into retrieval-ready Markdown and layout metadata.
- Price per page
- $1.50 / 1K pages
- Accepted documents
- PDF, PPT, JPEG as Base64
Resources
Top AI Models for Document Understanding
Specialist document models turn PDFs, presentations, scans, tables, and forms into usable text or fields. Compare cost per page, accepted files, output structure, layout evidence, language coverage, and deployment choices.
Showing 1–4 of 4 models
Mistral AI · Stable
Mistral OCR 4.1
Mistral's current Document AI OCR model for preserving reading order, tables, structure, locations, and confidence across multilingual files.
- Price per page
- $4 / 1K OCR pages · $5 / 1K annotated
- Accepted documents
- PDF, DOCX, PPTX, PNG, JPEG, AVIF and more
Google Cloud's Gemini-assisted parser for preserving document hierarchy and creating context-rich chunks for enterprise search and RAG.
- Price per page
- $10 / 1K pages, including initial chunking
- Accepted documents
- PDF, HTML, DOCX, PPTX, XLSX, XLSM
Google Cloud's Gemini-powered Document AI processor for extracting the exact fields and derived values defined in a business schema.
- Price per page
- $30 / 1K pages to 1M · $20 / 1K above
- Accepted documents
- PDF and image documents with a custom field schema
What to compare
- Price per page
- A current provider rate and the exact page or annotation unit it covers.
- Accepted documents
- The file and document types the listed endpoint is designed to process.
- What you get back
- Markdown, tables, chunks, fields, descriptions, or another machine-readable result.
How to use this library
Compare the job, not the hype.
A curated cross-category set of frontier, specialist, and category-defining models with practical API information for real product decisions.
Cody does not run a universal quality leaderboard. Provider claims are labeled, pricing excludes taxes and optional tools, and production buyers should verify the live provider terms before choosing a model.
- 01
Use first-party documentation for technical limits, lifecycle, pricing, regional access, and data policies whenever it exists.
- 02
Compare models only inside the same category and retain each provider's measurement basis.
- 03
Show Not published when a provider has not published a comparable fact instead of estimating it.
- 04
Treat prices and availability as dated snapshots and link every page to live provider documentation.
Speed keeps its units
Voice latency, text throughput, video render time, and world-model frame rate are not interchangeable.
Unknown stays unknown
A missing release date, country list, or latency number is labeled as unpublished instead of estimated.
Every fact has a date
Model pages link to the underlying source and show when pricing, access, and policies were last checked.
Put the model to work
Pick the model, then give it a better brief.
Use Cody's expanded prompt playbooks and curated agent skills to turn a model choice into a repeatable workflow.