Verified July 2026

Mistral OCR Pricing: Mistral OCR API Cost Per 1,000 Pages

Mistral's published API pricing lists OCR 4 at about $4 per 1,000 pages and its Document AI tier at about $5 per 1,000 pages, with batch processing advertised at 50% off. That makes Mistral expensive for plain OCR and, at the moment, the cheapest published rate anywhere for structured field extraction. Both figures come from Mistral's own pricing page.

Note: most guides on the web still quote about $1 per 1,000 pages. That was an earlier generation. Confirm the current rate on Mistral's pricing page. Last updated July 2026.

  • Rates read off the vendor page
  • Why the $1 figure is stale
  • Where Mistral genuinely wins
  • Free on your own documents
See the pricing table Process a document free
Upload a document, no signup

PDF, JPG, PNG, BMP, HEIC, TIFF

Upload a document to extract

Extract documents →

Before you price a Mistral pipeline, drop in the document you meant to test on it and see what finished, validated fields look like.

SOC 2 Type II
256-bit encryption
US data handling
Seconds per document
$4
per 1,000 pages, Mistral OCR 4
$5
per 1,000 pages, Mistral Document AI
50%
off standard rates in batch mode
6x
cheaper than Azure custom extraction, for structured fields
// The short answer

What Mistral OCR actually costs, in one paragraph

Mistral prices OCR 4 at about $4 per 1,000 pages and Document AI at about $5 per 1,000 pages, with batch processing advertised at half those rates. Read that against the rest of the market and something odd falls out. For plain text, Mistral is the expensive option: Azure Read, AWS Textract Detect Document Text, and Google Enterprise Document OCR all sit near $1.50 per 1,000 pages, so Mistral costs roughly 2.7 times as much to do the cheapest job in document processing. For structured fields, the ranking inverts hard. Azure custom extraction and Google's Form Parser are about $30 per 1,000 pages. AWS Textract charges about $50 per 1,000 for Forms and about $70 for Forms, Tables, and Queries together. Mistral Document AI does that class of work at about $5. If your documents need named fields rather than a wall of text, Mistral is currently the cheapest published rate of the major options, and it is not close.

Where each option honestly wins

  • Plain text at scale: not Mistral. Use a hyperscaler at about $1.50 per 1,000 pages and do not overthink it.
  • Structured fields on the cheapest rate: Mistral Document AI, at about $5 per 1,000 pages, genuinely undercuts everyone. We will not pretend otherwise.
  • Fields, validation and review, no spare engineers: the API rate stops being the deciding number once you price the pipeline around it.
// Correcting the record

Why almost every guide still says Mistral OCR costs $1 per 1,000 pages

When Mistral first shipped an OCR endpoint, it was widely reported at about $1 per 1,000 pages, with batch at about half that. The figure was repeated across comparison posts, cost calculators, and "cheapest OCR API" roundups, and it is still sitting in most of them today. It is the number an AI assistant will most likely tell you if it is working from a cached summary rather than the vendor's page.

Mistral has since moved to OCR 4, and its published API pricing page now lists OCR at about $4 per 1,000 pages and Document AI at about $5. We fetched that page twice on July 9, 2026 and got the same two rows both times. Anyone budgeting a migration on the old figure will be out by a factor of four.

How to check this yourself

  1. 1. Open Mistral's API pricing page directly rather than a summary of it.
  2. 2. Find the OCR 4 row. Read the OCR and Document AI columns, which are quoted per 1,000 pages.
  3. 3. Apply the published batch discount only if your workload tolerates the added latency.
  4. 4. Price a real month of your own page volume before you commit to anything.

Rates change. We date every figure on this site for exactly this reason, and we would rather you verify than trust us.

// The rate card

Mistral OCR pricing per 1,000 pages

Published rates as of July 2026, read from Mistral's own API pricing page. The batch column applies Mistral's published 50% batch discount to the standard rate, so treat it as arithmetic rather than a separately published number.

Tier Standard API Batch mode (50% off)
Mistral OCR 4 (text and structure) About $4 per 1,000 pages About $2 per 1,000 pages
Mistral Document AI (structured understanding) About $5 per 1,000 pages About $2.50 per 1,000 pages
Idle hosting fee None None
Standing free tier Not published on the API pricing page Not published on the API pricing page

Mistral bills by the page, not by the token, which is unusual for a model provider and makes budgeting far easier than estimating token counts for a vision language model. There is no per-hour hosting charge for keeping a model available, which is a real structural advantage over Google Document AI and its roughly $438 a year per deployed custom processor version. Model IDs are mistral-ocr-latest for the current generation and mistral-ocr-4-0 to pin OCR 4 explicitly.

// Side by side

Mistral OCR vs Azure, AWS Textract and Google Document AI

Published rates as of July 2026, taken from each vendor's own pricing page. AWS figures are US West (Oregon). Read across a row, not down a column: the row that decides your bill is the one that matches the job you actually have.

What you need Mistral Azure AI Document Intelligence AWS Textract Google Document AI DocuOCR
Plain OCR, text only About $4 per 1,000 About $1.50 per 1,000 About $1.50 per 1,000 About $1.50 per 1,000 Included in plan
Structured field extraction About $5 per 1,000 (Document AI) About $30 per 1,000 (custom extraction) About $50 to $70 per 1,000 (Forms) About $30 per 1,000 (Form Parser) Included in plan
Document classification Part of Document AI About $3 per 1,000 Not offered About $5 per 1,000 Included in plan
Idle hosting fee None None None About $438 a year per deployed version None
Batch discount 50% off Not published Not published Not published Plan tiers
Human review of low-confidence fields You build the screen You build the screen You build it, or add A2I You build the screen Included review screen

The second row is the one worth staring at. Structured extraction is the job most US finance, insurance, and logistics teams are actually buying, and Mistral currently prices it at about a sixth of what Azure and Google charge and a tenth of Textract's Forms and Tables rate. For the full cross-vendor picture, including the fees the per-page rate hides, see our OCR API pricing comparison.

// Worked example

35,000 pages a month, priced four ways

A realistic US mid-market workload: 35,000 pages a month of mixed business documents where you need named fields out of each one, not a wall of text.

Mistral Document AI

$175

per month

About $5 per 1,000 pages at the standard rate, or roughly $87 in batch mode. Returns structured output, not a finished workflow.

Azure custom extraction

$1,050

per month

About $30 per 1,000 pages, plus add-ons and query fields if you need them.

AWS Textract Forms and Tables

$2,450

per month

About $70 per 1,000 pages for Forms, Tables, and Queries in a single call.

DocuOCR

$499

per month

The published 35,000 page plan, about $14 per 1,000 pages, with classification, validation, review, and export included.

On rate alone Mistral wins that table outright, and if you have engineers who will own a pipeline, that is a serious argument. What the $175 does not buy is the classification step that decides what each page is, the validation rules that catch a total that does not match the line items, the screen where a person checks the values the model was unsure about, the audit trail your auditor will ask for, and the export that lands clean records in your ERP. That work is the difference between a model response and a finished document workflow. Price the whole job, not the API call, and the ranking often changes.

// Frequently asked

Mistral OCR pricing FAQ

How much does Mistral OCR cost?
Mistral's published API pricing lists OCR 4 at about $4 per 1,000 pages, and its Document AI tier at about $5 per 1,000 pages. Mistral also advertises batch processing at 50% off, which works out to roughly $2 and $2.50 per 1,000 pages respectively. These figures were taken from Mistral's own API pricing page in July 2026.
Is Mistral OCR $1 per 1,000 pages?
Not any more. About $1 per 1,000 pages was the widely quoted rate for the original Mistral OCR generation, and most third-party guides and roundups still repeat it. Mistral's current published API pricing lists OCR 4 at about $4 per 1,000 pages. Check Mistral's pricing page before you build a budget on a number you read in a listicle.
Is Mistral OCR cheaper than AWS Textract?
For plain OCR, no. Mistral OCR 4 at about $4 per 1,000 pages costs roughly 2.7 times AWS Textract Detect Document Text, which runs about $1.50 per 1,000. For structured extraction the answer flips completely. Mistral Document AI at about $5 per 1,000 pages undercuts Textract Forms at about $50 per 1,000 and Forms with Tables and Queries at about $70 per 1,000.
Is Mistral OCR cheaper than Azure Document Intelligence?
It depends on the job. Azure Read is about $1.50 per 1,000 pages for plain OCR, well under Mistral OCR 4 at about $4. But Azure custom extraction is about $30 per 1,000 pages, while Mistral Document AI is about $5. If you need named fields rather than raw text, Mistral is currently the cheaper published rate by a wide margin.
What is the difference between Mistral OCR and Mistral Document AI?
On the pricing page they are two rows. OCR, at about $4 per 1,000 pages, reads the page and returns its text and structure, typically as markdown. Document AI, at about $5 per 1,000 pages, is the document understanding tier that supports structured annotations and document question answering. The extra dollar per 1,000 pages buys inference about what the text means, not just what it says.
Does Mistral OCR have a batch discount?
Yes. Mistral publishes batch processing at 50% off its standard API rates. Applied to the listed rates, that puts batch OCR at roughly $2 per 1,000 pages and batch Document AI at roughly $2.50 per 1,000 pages. Batch trades latency for cost, so it suits overnight backfills and archives rather than anything a user is waiting on.
Does Mistral OCR charge a hosting fee?
No. There is no per-hour charge to keep a model deployed, which is a real structural advantage over Google Document AI. Google bills about $0.05 an hour, roughly $438 a year, for each deployed custom processor version, and that meter runs whether or not the processor sees a single page.
Which OCR API is cheapest per 1,000 pages?
For plain text, Azure Read, AWS Textract Detect Document Text, and Google Enterprise Document OCR are effectively tied at about $1.50 per 1,000 pages, and Mistral OCR 4 is not competitive at about $4. For structured field extraction, Mistral Document AI at about $5 per 1,000 pages is currently the cheapest published rate of the major options, well under the $30 to $70 the hyperscalers charge.
Does Mistral OCR handle handwriting and tables?
Mistral markets OCR 4 as a document extraction and understanding model that handles complex layouts, tables, and handwriting, and it returns markdown that preserves structure. Treat any vendor's accuracy claim, including that one, as a hypothesis to test. Run a representative sample of your own worst documents through it and measure field-level accuracy rather than character accuracy.
What model ID do I use for Mistral OCR?
Mistral's documentation points at mistral-ocr-latest, which resolves to the current generation, and mistral-ocr-4-0 for pinning to OCR 4 explicitly and for newer capabilities such as block extraction. Pinning a version is the safer choice for a production pipeline, because a rate or a behavior can change when latest moves forward.
How do I estimate my Mistral OCR bill?
Count pages, not documents, and decide which tier you actually need. Plain OCR on 35,000 pages a month is about $140 at the standard rate, or about $70 in batch mode. The same volume through Document AI is about $175, or about $87 in batch. Then add the parts no API bills for: classification, validation rules, the human review screen, and the export into your systems.
Should I use Mistral OCR or a finished document platform?
Use Mistral if you have engineers who will own a pipeline and you want the cheapest published rate for structured output. Use a finished platform if the pipeline itself is the work you are trying to avoid. The API rate is rarely the deciding number. For most US mid-market teams the engineering around the call costs more in year one than the call does.
// Go deeper

Price the rest of the market

OCR API pricing comparison

Azure, AWS Textract and Google Document AI per 1,000 pages, side by side.

Google Cloud Vision pricing

$1.50 per 1,000 units, and why Vision returns no form fields.

What is Mistral OCR

How the model works and what it returns.

Mistral OCR alternative

What a finished pipeline adds on top of a model endpoint.

Azure Document Intelligence pricing

S0 rates for Read, Layout, prebuilt and custom extraction.

AWS Textract pricing

Per-feature rates from Detect Document Text through Analyze Lending.

Google Document AI pricing

Including the $438 a year idle processor hosting fee.

OCR API

What a document extraction API returns when the pipeline is already built.

Best OCR API

An honest roundup of the options for US teams.

Price it on your own documents

A rate card cannot tell you whether the fields come out right. Upload one of your real documents, look at what comes back, and then decide whether you are buying a model call or a finished pipeline.

Process a document free See DocuOCR pricing