# Mistral OCR vs AWS Textract

Updated: 2026-10-16

## Verdict

**Choose Mistral OCR when** you want Markdown out of the box with block-level bounding boxes and confidence at $4 per 1,000 pages ($5 annotated; Mistral stated a 50 percent batch discount for OCR 4, not re-confirmed for 4.1), schema extraction through annotations in the same call, and handwriting support, including cursive, per Mistral's OCR 3 announcement (no handwriting accuracy figure published). Files are capped at 50 MB and 1,000 pages, and Mistral's benchmarks are self-published with competitor numbers it reproduced internally.

**Choose AWS Textract when** you already run on AWS and need JSON blocks with geometry and confidence, forms, tables, queries, signatures, invoices, and IDs priced per feature ($1.50 per 1,000 pages for text, $4 for layout, $15 for tables or queries, $50 for forms), with HIPAA eligibility, PrivateLink, and GovCloud. Handwriting and queries are English only, and no native Markdown output is documented.

**Choose Nutrient Data Extraction when** you need OCR plus Markdown or spatial JSON in 100-plus languages, a mode ladder from 1.5 credits per page with OCR, an Auto mode that picks the mode per page, and grounded schema extraction with match labels, with 5,000 credits renewing monthly. Nutrient was founded as PSPDFKit in 2011. 3,000-plus organizations, including 15 percent of the Global 500, rely on it, with over a billion document interactions a year. The same extraction engine runs hosted, self-hosted in the Python and Java SDKs with no external API calls for OCR and ICR, or on a Mac in Extract Studio.

## At a glance

_Mistral OCR / Document AI, AWS Textract, Nutrient Data Extraction API; facts fetched 2026-10-15 to 2026-10-16_

| Fact | Mistral OCR / Document AI | AWS Textract | Nutrient Data Extraction API |
| --- | --- | --- | --- |
| Product type | hosted API (single /v1/ocr endpoint); Document AI adds structured-output parameters to the same call; self-hosting offered to enterprise customers [1] | hosted API (AWS service; sync and async operations) [2] | Hosted Data Extraction API, web Studio, and macOS Extract Studio [3] |
| Output formats | JSON (pages[] with markdown, images, tables, hyperlinks, header, footer, dimensions, confidence_scores, blocks), Markdown (per page), HTML tables (table_format=html), Markdown tables (table_format=markdown) [4] | JSON (Block objects: PAGE, LINE, WORD, TABLE, CELL, KEY_VALUE_SET, QUERY, LAYOUT_* etc.) [5] | Spatial JSON, Markdown, Schema-shaped JSON with citations, Ranked labels with independent 0-1 scores, Split page ranges and ZIP files [6] |
| OCR for scans | Yes [7] | Yes [8] | Yes: structure, understand, and agentic; text has no OCR; 100+ languages [9] |
| Handwriting | Yes [7] | English only [10] | Yes: printed-style in understand; cursive/freeform needs agentic [11] |
| Tables | Yes [4] | Yes [12] | Yes: spans and bounds; complex tables may need agentic mode [11] |
| Layout and bounding boxes | Yes [4] | Yes [5] | Yes: element and optional word bounds; word bounds exclude XLSX [11] |
| Per-field confidence | OCR confidence at page, block, or word level; per-extracted-field confidence for annotations was not documented in the sources reviewed [4] | Yes [8] | Yes: 0-1 relative signal, not a probability [13] |
| Source citations and grounding | Partly [1] | Partly [14] | Yes: field boxes, match labels, confidence, and source boxes [13] |
| Schema extraction | Yes [15] | Partly [10] | Yes: closed JSON Schema subset; 128 KB, 2,000 nodes, depth 5 [16] |
| Classification | Partly [15] | Partly [17] | Yes: zero-shot, 2+ labels; independent 0-1 scores; 1 credit/page [18] |
| Document splitting | Not published [4] | Partly [17] | Yes: boundary detection for PDF/images; 3 credits/page [19] |
| Free tier | Not published [20] | Free Tier lasts three months for new AWS customers: Detect Document Text 1,000 pages/month; Analyze Document: 1,000 pages/month Signatures only, 100 pages/month Forms/Tables/Layout, 100 pages/month each for Queries, Forms+Queries, Tables+Queries, Forms+Tables+Queries; no free tier for Custom Queries; AnalyzeExpense 100, AnalyzeID 100, Analyze Lending 2,000 pages/month [21] | 5,000 credits/month; no card; all parse modes, including Auto [22] |
| Pricing model | per 1,000 pages (OCR 4.1: $4; annotated pages $5); Mistral published a 50 percent batch discount for OCR 4 that was not reconfirmed for 4.1 [23] | per page, by API/feature combination; volume tier above 1M pages/month; separate prices per region [21] | Credits/page by endpoint and mode; Auto routes each page and bills that mode's rate; subscription plus capped overage [24] |
| List price examples | OCR 4.1 (mistral-ocr-4-1 / mistral-ocr-latest) API: $4 per 1,000 pages (OCR + bounding boxes, block labels, confidence scores)<br>OCR 4.1 Annotated pages / Document AI: $5 per 1,000 pages (adds structured JSON annotation)<br>OCR 4 Batch API: $2 per 1,000 pages (50% discount) (batch processing; stated in OCR 4 blog for OCR 4)<br>Historic: OCR 3 (mistral-ocr-2512), Dec 2025: $2 per 1,000 pages; annotations $3 per 1,000; batch $1 per 1,000 (superseded; see https://mistral.ai/news/mistral-ocr-3/)<br>Historic: Mistral OCR (2503), Mar 2025: 1000 pages / $ (about $1 per 1,000 pages); batch roughly double pages per dollar (superseded; see https://mistral.ai/news/mistral-ocr/) [1] | Detect Document Text: $0.0015/page ($1.50 per 1,000); $0.0006 over 1M (printed text + handwriting OCR)<br>Analyze Document: Layout: $0.004/page ($4 per 1,000); $0.003 over 1M (layout elements; included free with any combo of Forms/Tables/Queries)<br>Analyze Document: Tables: $0.015/page ($15 per 1,000); $0.010 over 1M (tables + text (Layout included))<br>Analyze Document: Forms: $0.05/page ($50 per 1,000); $0.04 over 1M (key-value pairs + text (Layout included))<br>Analyze Document: Queries: $0.015/page ($15 per 1,000); $0.010 over 1M (natural-language queries (Layout included))<br>Analyze Document: Tables + Forms: $0.065/page (sum of $0.015 + $0.05 per page example) (tables + forms)<br>Analyze Document: Forms + Queries: $0.055/page; $0.045 over 1M ()<br>Analyze Document: Tables + Queries: $0.020/page; $0.015 over 1M ()<br>Analyze Document: Tables + Forms + Queries: $0.070/page; $0.055 over 1M ()<br>Analyze Document: Signatures: $0.0035/page; $0.0014 over 1M (signature detection (free with other feature combos))<br>Custom Queries (adapter): $0.025/page; $0.015 over 1M (adapter training is free; +Tables $0.030, +Forms $0.065, +Forms+Tables $0.080)<br>Analyze Expense: $0.010/page; $0.008 over 1M (invoices/receipts)<br>Analyze ID: $0.025/ID; $0.010 over 100K (US passports and driver's licenses)<br>Analyze Lending: $0.070/page; $0.055 over 1M (classify + split + extract mortgage docs) [25] | Per-page credits: parse 1-18; extract 7-24 (classify 1; split 3)<br>Auto mode: bills each page at the mode it selects (recommended starting mode; routes clean pages to structure and scans or complex forms to understand)<br>Free: $0/month (5,000 credits; PAYG after a $25 wallet load)<br>Starter: $59/month; $49/month annual (25,000 credits; $0.0028 per extra credit)<br>Pro: $500/month; $450/month annual (500,000 credits; $0.0012 per extra credit) [22] |
| Deployment | cloud (Mistral Studio API), Amazon SageMaker, Microsoft Foundry, Snowflake Parse Document (listed 'coming soon' in Jun 2026), self-hosted single container (enterprise customers) [1] | cloud only (AWS), commercial regions: us-east-1, us-east-2, us-west-1, us-west-2, ca-central-1, eu-central-1, eu-west-1, eu-west-2, eu-west-3, eu-south-2, ap-south-1, ap-northeast-2, ap-southeast-1, ap-southeast-2, AWS GovCloud (US-East, US-West), FIPS endpoints in some US/Canada regions, VPC endpoints via AWS PrivateLink [26] | Hosted API: Nutrient-managed cloud, Self-hosted: Data Extraction SDK for Python and Java, no network required; OCR and ICR run on your infrastructure with zero external API calls, Local VLM: Ollama, LM Studio, or vLLM on your own hardware, or your own OpenAI or Anthropic key, Desktop: Extract Studio for Apple silicon Macs (macOS Tahoe 26.4+); the engine runs on the Mac and, with a local model, documents never leave it; Nutrient counts pages for credits, .NET data extraction SDK also offered [27] |
| SDKs | Python (mistralai), TypeScript, REST/curl [4] | AWS SDKs (Python, Java, JavaScript listed on product page), AWS CLI, REST API [28] | REST API, Python SDK, Java SDK, .NET page referenced, curl, JavaScript, and raw HTTP samples [29] |
| MCP or agent tools | Not published [4] | Not published [8] | Hosted MCP, Data Extraction Studio, and agent skills [30] |
| Maximum file size or pages | Uploaded document files must be 50 MB or less and contain 1,000 pages or fewer [4] | Sync: JPEG/PNG/PDF/TIFF up to 10 MB, PDF/TIFF limited to 1 page. Async: JPEG/PNG up to 10 MB; PDF/TIFF up to 500 MB and 3,000 pages. Images up to 10,000 px per side; PDF max 40 in / 9,000 pt [10] | Direct upload: 150 MB, URL input: 50 MB, URL timeout: 30 seconds, Processing: 5 minutes; HTTP 408 if exceeded, Page limit: not published, Test-key limits: lower, unpublished; Support-enabled [31] |
| Compliance | Not published [32] | HIPAA eligible (BAA required), PCI, ISO, SOC, VPC endpoints via PrivateLink, FIPS endpoints in some regions [8] | SOC 2: Type 2, audited annually, BAA: no public statement, Transport: HTTPS/TLS only, No-retention plans: files discarded after processing, Retention plans: runs stored; may train/evaluate models, Payments: Paddle [33] |
| Public benchmarks | OlmOCRBench 85.20 (overall score; Mistral reports top among systems it tested (competitor scores are Mistral internal reproductions)), Mistral AI, 2026-06-23 ([source](https://mistral.ai/news/ocr-4/))<br>OmniDocBench 93.07 (Mistral OCR 4 score; Mistral flags known scoring artifacts and calls aggregate scores 'directional'), Mistral AI, 2026-06-23 ([source](https://mistral.ai/news/ocr-4/))<br>Crawl Multilingual (internal) .98 (OCR 4 internal multilingual eval), Mistral AI, 2026-06-23 ([source](https://mistral.ai/news/ocr-4/))<br>Human preference (600+ docs, 12+ languages, blind annotators) 72% average win rate for OCR 4 across systems tested (pairwise preference), Mistral AI, 2026-06-23 ([source](https://mistral.ai/news/ocr-4/))<br>Internal benchmarks on forms, scans, tables, handwriting 74% overall win rate of OCR 3 over OCR 2 (win rate vs previous Mistral model), Mistral AI, 2025-12-17 ([source](https://mistral.ai/news/mistral-ocr-3/))<br>Mistral OCR launch internal text-only test set (Overall / Math / Multilingual / Scanned / Tables) Mistral OCR 2503 94.89 / 94.29 / 89.55 / 98.96 / 96.12; Google Document AI 83.42 / 80.29 / 86.42 / 92.77 / 78.16; Azure OCR 89.52 / 85.72 / 87.52 / 94.65 / 89.52; Gemini-2.0-Flash-001 88.69 / 84.18 / 85.80 / 95.11 / 91.46; GPT-4o-2024-11-20 89.77 / 87.55 / 86.00 / 94.58 / 91.70 (accuracy (fuzzy match), internal test set), Mistral AI, 2025-03-06 ([source](https://mistral.ai/news/mistral-ocr/))<br>Mistral OCR launch per-language fuzzy match (overall multilingual) Mistral OCR 2503 99.02; Azure OCR 97.31; Gemini-2.0-Flash-001 96.53; Google Document AI 95.88 (fuzzy match in generation), Mistral AI, 2025-03-06 ([source](https://mistral.ai/news/mistral-ocr/)) [1] |  [8] | ParseBench public row 76.4; paired nightly: 82.8 (overall 0 to 100), LlamaIndex, 2026-10-15 ([source](https://parsebench.ai/))<br>ParseBench nightly 82.8; paired public row: 76.4 (overall 0 to 100), Nutrient, 2026-10-15 ([source](https://www.nutrient.io/api/data-extraction-api/benchmarks/))<br>opendataloader-bench, Nutrient's own run understand 0.933; agentic 0.930; structure 0.892; text 0.889 (overall 0 to 1), Nutrient, 2026-10-16 ([source](https://www.nutrient.io/api/data-extraction-api/benchmarks/))<br>opendataloader.org leaderboard Nutrient row (free CLI) 0.885; opendataloader hybrid 0.907 (overall 0 to 1), opendataloader.org, 2026-09-23 ([source](https://opendataloader.org/docs/benchmark)) [35] |

## Output and languages

Mistral returns per-page Markdown with images, tables as Markdown or HTML, blocks with labels and bounding boxes, and confidence at page, block, or word granularity. Textract returns JSON blocks (page, line, word, table, cell, key-value, query, layout) with geometry and 0 to 100 confidence; printed text in six languages, handwriting in English only. Nutrient returns Markdown or spatial elements with bounds, reading order, and OCR confidence, in more than 100 OCR languages, with printed-style handwriting in understand mode and cursive in agentic mode.

## Pricing per 1,000 pages

Mistral OCR 4.1: $4 per 1,000 pages, $5 per 1,000 annotated pages; Mistral stated a 50 percent batch discount for OCR 4, not re-confirmed for 4.1. Textract (US East): Detect Document Text $1.50, Layout $4, Tables $15, Queries $15, Forms $50, Signatures $3.50, Expense $10, ID $25 per 1,000 IDs; features stack when combined, and prices drop above a million pages a month. Textract's free tier runs three months with 1,000 pages a month of text detection and 100 pages a month for forms, tables, and layout.

Nutrient: structure mode (OCR, basic segmentation) 1.5 credits per page, understand 9, agentic 18; extract adds 6; the free plan renews 5,000 credits a month.

## Structured extraction

Mistral's annotations take a JSON schema and run a second model call over the OCR output. Textract's Queries answer up to 15 natural-language questions per page synchronously (30 asynchronously), in English, and custom queries need a trained adapter. Nutrient's `/extraction/extract` takes an inline JSON Schema subset (no `$ref` or `anyOf`, depth 5) and returns per-field bounding boxes, match labels, and relative confidence.

## Limits and compliance

Mistral: 50 MB and 1,000 pages per file; rate limits per organization in the admin panel; its trust center could not be read, so no certification is stated here. Textract: sync calls take one page and 10 MB; async handles 500 MB and 3,000 pages via S3; HIPAA eligible with a BAA, PCI, ISO, SOC, FIPS endpoints in some regions, GovCloud. Nutrient: 150 MB direct upload, 50 MB by URL, SOC 2 Type 2 audited controls, HTTPS only; no BAA statement found on public pages.

## Published scores, with who ran them

Numbers here were run by the vendors named, several of them in the table above. Read them as directional and test on your own documents.

Nutrient's published benchmark results, as recorded on 16 October 2026: on LlamaIndex's ParseBench the public leaderboard row for Nutrient agentic is 76.4. Nutrient's own nightly run of 15 October 2026 scores 82.8 on the same test and is not on parsebench.ai yet; in that nightly run, understand mode scores 64.0. LlamaParse Agentic Plus leads the public leaderboard at 90.2. Nutrient is not on Datalab's OmniParseBench or OmniExtractBench public leaderboards. In Nutrient's own run on the opendataloader-bench corpus the API scores 0.933 in understand mode and 0.930 in agentic; on the public opendataloader.org leaderboard Nutrient's row (the free CLI) scores 0.885, behind opendataloader hybrid at 0.907.

## Where Nutrient differs

The other products on this page are parsers or extraction APIs. Nutrient's Data Extraction API is the same kind of product, built by a company that has shipped document SDKs to enterprises since 2011. The facts below come from Nutrient's own public pages.

- Founded as PSPDFKit in 2011, and the SDKs developers knew as PSPDFKit continue as Nutrient SDKs. Thousands of organizations in 80-plus countries, including Fortune 500s and governments, use them ([about page](https://www.nutrient.io/company/about/)).
- 3,000-plus organizations and 15 percent of Global 500 companies rely on Nutrient, with over a billion document interactions a year ([nutrient.io](https://www.nutrient.io/)).
- One engine, four deployments: hosted API in Nutrient's cloud; the Python and Java Data Extraction SDKs, self-hosted with no network required and no external API calls for OCR and ICR, with a local VLM through Ollama, LM Studio, or vLLM ([Python SDK](https://www.nutrient.io/sdk/python/pdf-data-extraction/)); Extract Studio on Apple silicon, where documents never leave the Mac when the model runs locally ([Extract Studio](https://www.nutrient.io/api/data-extraction-api/macos-desktop-app/)); and the wider platform in the browser with WebAssembly, on your own servers with Document Engine, or air-gapped ([nutrient.io](https://www.nutrient.io/)).
- Extraction is one credit pool with parse, classify, and split, and every extracted field carries a bounding box and a match label, so a reviewer can check the source instead of trusting the model. SOC 2 Type 2 audited.

## What to test before choosing

- Run the same 20 documents through each API: five clean born-digital PDFs, five scans, five with tables that span pages, five with the fields your downstream system needs. Score reading order, table cells, and the fields you will actually consume.
- Count the per-page cost at the mode or tier that passed your sample, not the cheapest tier on the pricing page. Every vendor here prices accuracy upward.
- Check what the API returns when it is unsure: a confidence number, a bounding box, a `not_found` label, or nothing. That decides whether you can build a review queue.
- Confirm retention and deployment in writing for the plan you will buy. Free and pay-as-you-go tiers often exclude BAAs, zero-retention, and private deployment.

## Frequently asked questions

### Is Mistral OCR better than Textract?

Mistral's launch table (March 2025, a superseded model) compared itself with Azure OCR and Google Document AI but not Textract, and its later OlmOCRBench 85.20 and OmniDocBench 93.07 figures are self-published with internally reproduced competitor scores. No accuracy numbers were found in Textract's docs or FAQ. Run both on your own documents.

### Does Textract output Markdown?

No native Markdown output was found in Textract's documentation; it returns JSON blocks, and tables can be exported to CSV with sample code. Mistral returns Markdown per page. Nutrient returns Markdown or spatial JSON.

### Which handles handwriting?

Textract recognizes handwriting in English only. Mistral claims cursive and handwriting over printed forms in its OCR 3 and OCR 4 announcements. Nutrient reads printed-style handwriting in understand mode and recommends agentic mode for cursive, and its guide warns a recognition can be confident but wrong.

### Which is cheapest for plain text OCR?

Textract Detect Document Text lists $1.50 per 1,000 pages; Mistral OCR 4.1 lists $4. Mistral published a $2 batch rate for OCR 4, but we did not reconfirm it for 4.1. Nutrient structure mode is 1.5 credits per page: $4.20 per 1,000 pages at the Starter rate or $1.80 at the Pro rate (our arithmetic).

### Can I run either on-prem?

Textract is cloud only on AWS. Mistral offers a self-hosted single container to enterprise customers and runs on SageMaker and Microsoft Foundry. Nutrient runs the same engine three ways: the hosted API; self-hosted in the Python and Java Data Extraction SDKs, where OCR and ICR run on your infrastructure with no external API calls and the VLM stage can run locally through Ollama, LM Studio, or vLLM; and on a Mac in Extract Studio, where documents never leave the machine when the model runs locally.

## Sources

1. [https://mistral.ai/news/ocr-4/](https://mistral.ai/news/ocr-4/) — fetched 2026-10-15
2. [https://docs.aws.amazon.com/textract/latest/dg/what-is.html](https://docs.aws.amazon.com/textract/latest/dg/what-is.html) — fetched 2026-10-15
3. [https://www.nutrient.io/guides/dws-data-extraction/api-overview/](https://www.nutrient.io/guides/dws-data-extraction/api-overview/) — fetched 2026-10-15
4. [https://docs.mistral.ai/studio/document-processing/basic_ocr](https://docs.mistral.ai/studio/document-processing/basic_ocr) — fetched 2026-10-15
5. [https://docs.aws.amazon.com/textract/latest/dg/layoutresponse.html](https://docs.aws.amazon.com/textract/latest/dg/layoutresponse.html) — fetched 2026-10-15
6. [https://www.nutrient.io/guides/dws-data-extraction/parsing/](https://www.nutrient.io/guides/dws-data-extraction/parsing/) — fetched 2026-10-15
7. [https://mistral.ai/news/mistral-ocr-3/](https://mistral.ai/news/mistral-ocr-3/) — fetched 2026-10-15
8. [https://aws.amazon.com/textract/faqs/](https://aws.amazon.com/textract/faqs/) — fetched 2026-10-15
9. [https://www.nutrient.io/guides/dws-data-extraction/parsing/processing-modes/](https://www.nutrient.io/guides/dws-data-extraction/parsing/processing-modes/) — fetched 2026-10-15
10. [https://docs.aws.amazon.com/textract/latest/dg/limits-document.html](https://docs.aws.amazon.com/textract/latest/dg/limits-document.html) — fetched 2026-10-15
11. [https://www.nutrient.io/guides/dws-data-extraction/parsing/extract-document-elements/](https://www.nutrient.io/guides/dws-data-extraction/parsing/extract-document-elements/) — fetched 2026-10-15
12. [https://docs.aws.amazon.com/textract/latest/dg/how-it-works-tables.html](https://docs.aws.amazon.com/textract/latest/dg/how-it-works-tables.html) — fetched 2026-10-15
13. [https://www.nutrient.io/guides/dws-data-extraction/extract/citations-and-confidence/](https://www.nutrient.io/guides/dws-data-extraction/extract/citations-and-confidence/) — fetched 2026-10-15
14. [https://docs.aws.amazon.com/textract/latest/dg/queryresponse.html](https://docs.aws.amazon.com/textract/latest/dg/queryresponse.html) — fetched 2026-10-15
15. [https://docs.mistral.ai/capabilities/document_ai/annotations/](https://docs.mistral.ai/capabilities/document_ai/annotations/) — fetched 2026-10-15
16. [https://www.nutrient.io/guides/dws-data-extraction/extract/define-a-schema/](https://www.nutrient.io/guides/dws-data-extraction/extract/define-a-schema/) — fetched 2026-10-15
17. [https://docs.aws.amazon.com/textract/latest/dg/lending-document-classification-extraction.html](https://docs.aws.amazon.com/textract/latest/dg/lending-document-classification-extraction.html) — fetched 2026-10-15
18. [https://www.nutrient.io/guides/dws-data-extraction/classify/](https://www.nutrient.io/guides/dws-data-extraction/classify/) — fetched 2026-10-15
19. [https://www.nutrient.io/api/data-extraction-api/split/](https://www.nutrient.io/api/data-extraction-api/split/) — fetched 2026-10-15
20. [https://docs.mistral.ai/admin/user-management-finops/tier](https://docs.mistral.ai/admin/user-management-finops/tier) — fetched 2026-10-15
21. [https://aws.amazon.com/textract/pricing/](https://aws.amazon.com/textract/pricing/) — fetched 2026-10-15
22. [https://www.nutrient.io/api/pricing/data-extraction-api/](https://www.nutrient.io/api/pricing/data-extraction-api/) — fetched 2026-10-15
23. [https://docs.mistral.ai/models/ocr-4-1](https://docs.mistral.ai/models/ocr-4-1) — fetched 2026-10-15
24. [https://www.nutrient.io/guides/dws-data-extraction/pricing/](https://www.nutrient.io/guides/dws-data-extraction/pricing/) — fetched 2026-10-15
25. [https://b0.p.awsstatic.com/pricing/2.0/meteredUnitMaps/textract/USD/current/textract.json](https://b0.p.awsstatic.com/pricing/2.0/meteredUnitMaps/textract/USD/current/textract.json) — fetched 2026-10-15
26. [https://docs.aws.amazon.com/general/latest/gr/textract.html](https://docs.aws.amazon.com/general/latest/gr/textract.html) — fetched 2026-10-15
27. [https://www.nutrient.io/sdk/python/pdf-data-extraction/](https://www.nutrient.io/sdk/python/pdf-data-extraction/) — fetched 2026-10-16
28. [https://aws.amazon.com/textract/](https://aws.amazon.com/textract/) — fetched 2026-10-15
29. [https://www.nutrient.io/guides/dws-data-extraction/getting-started/](https://www.nutrient.io/guides/dws-data-extraction/getting-started/) — fetched 2026-10-15
30. [https://www.nutrient.io/api/mcp-server-pdf-automation-llm/](https://www.nutrient.io/api/mcp-server-pdf-automation-llm/) — fetched 2026-10-15
31. [https://www.nutrient.io/guides/dws-data-extraction/file-types/](https://www.nutrient.io/guides/dws-data-extraction/file-types/) — fetched 2026-10-15
32. [https://trust.mistral.ai/](https://trust.mistral.ai/) — fetched 2026-10-15
33. [https://www.nutrient.io/guides/dws-data-extraction/security/](https://www.nutrient.io/guides/dws-data-extraction/security/) — fetched 2026-10-15
34. [https://mistral.ai/news/mistral-ocr/](https://mistral.ai/news/mistral-ocr/) — fetched 2026-10-15
35. [https://www.nutrient.io/api/data-extraction-api/benchmarks/](https://www.nutrient.io/api/data-extraction-api/benchmarks/) — fetched 2026-10-15
36. [https://parsebench.ai/](https://parsebench.ai/) — fetched 2026-10-15
37. [https://opendataloader.org/docs/benchmark](https://opendataloader.org/docs/benchmark) — fetched 2026-10-15