Mistral OCR 4
Mistral's June 2026 document OCR model, retired in September 2026 and replaced by OCR 4.1
Mistral AI has replaced this model with Mistral OCR 4.1.
Context window
16K
Input / 1M tokens
$4.00 per 1,000 pages (API) / $2.00 per 1,000 pages (Batch)
Output / 1M tokens
$5.00 per 1,000 pages (Document AI annotation tier)
Provider
Mistral AI
Billed per page, not per token: $4 per 1,000 pages via the standard API, $2 per 1,000 pages with the Batch API, and $5 per 1,000 annotated pages through the Document AI layer. Mistral deprecated OCR 4.0 on September 29, 2026 and retired it on September 30, 2026; OCR 4.1 costs the same. · Data verified 2026-11-11
Mistral OCR 4 (API name mistral-ocr-4-0) is Mistral AI's document-understanding and OCR model from June 23, 2026. It extracts text with bounding boxes, classifies typed blocks (titles, tables, equations, signatures), attaches per-page and per-word confidence scores, and outputs clean Markdown. It accepts PDF, DOC, PPT, and OpenDocument files, supports 170 languages across 10 language groups, and offers an optional Document AI layer for reshaping output to a custom JSON schema. OCR 4.1 replaced it: Mistral retired OCR 4.0 from the API on September 30, 2026, and the mistral-ocr-4 and mistral-ocr-latest aliases now point to OCR 4.1 at the same per-page price.
Capability index
Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.
How to access it
Retired from the Mistral API on September 30, 2026. Use OCR 4.1 instead: the mistral-ocr-4 and mistral-ocr-latest aliases now point to it (pinned ID mistral-ocr-4-1), at the same $4 per 1,000 pages.
Strengths
- ✓Top-tier OCR accuracy (OlmOCRBench 85.20, OmniDocBench 93.07)
- ✓Structured Markdown output with typed block classification and bounding boxes
- ✓Per-page and per-word confidence scores for reliable downstream use
- ✓170-language coverage across 10 language groups, including low-resource languages
- ✓Extremely low per-page cost and optional fully self-hosted deployment
Best for developers who...
When to choose it (and when not to)
Reach for Mistral OCR 4 when...
- →You maintain code written for OCR 4: the same calls through mistral-ocr-latest now reach OCR 4.1
- →You need PDFs, scans or office files turned into structured Markdown (use OCR 4.1, its replacement)
Look elsewhere if...
- ✕Any new integration (OCR 4.0 was retired on September 30, 2026; use OCR 4.1)
- ✕You need a general-purpose chat or reasoning model (this is OCR/document-only)
- ✕Your input is plain text with no document/image structure to extract
- ✕You need conversational tool-use or code generation
How to use it
- ›Send high-resolution scans/PDFs for best extraction accuracy
- ›Use the Document AI layer with a JSON schema when you need typed, structured fields
- ›Use the Batch API to halve cost ($2 vs $4 per 1,000 pages) for large jobs
- ›Inspect per-word/per-page confidence scores to flag low-confidence regions for review
Quickstart
Pythonfrom mistralai import Mistral
client = Mistral(api_key="YOUR_MISTRAL_API_KEY")
resp = client.ocr.process(
model="mistral-ocr-latest",
document={
"type": "document_url",
"document_url": "https://example.com/invoice.pdf",
},
)
print(resp.pages[0].markdown)Uses the official 'mistralai' SDK OCR endpoint. The mistral-ocr-latest alias now resolves to OCR 4.1, because OCR 4.0 was retired on September 30, 2026. Billing is per page processed, not per token.
API model id: mistral-ocr-4-0
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| OlmOCRBench | 85.20 | Top score reported in Mistral's OCR 4 announcement. |
| OmniDocBench | 93.07 | Reported in Mistral's OCR 4 announcement. |
| Human preference (head-to-head) | 72% win rate | 72% win rate against competing OCR systems in human preference evaluations, per Mistral. |
| Human evaluation (blind) | 72% | Independent annotators preferred OCR 4 in 72% of comparisons across 600+ documents |
| Human Evaluation (Blind Test) | 72% | Independent annotators preferred OCR 4 over every competing system tested across 600+ documents in 12+ languages |
| Human Evaluation | 72% win rate | Win rate in blind human evaluations vs. leading competitors across 600+ documents |
| Human Evaluation (Independent Annotators) | 72% | Win rate across 600+ documents in 12+ languages |
| Crawl Multilingual | 0.98 | Internal Mistral evaluation, leads on 8 of 10 language groups |
| Human Evaluation (Blind) | 72% | Win rate across 600+ documents in 12+ languages |
| Human Preference Win Rate | 72% | Blind evaluation across 600+ documents in 12+ languages |
| Human Evaluation Win Rate | 72% | 72% average win rate in blind human evaluations across 600+ documents in 12+ languages |
| Human Preference (Blind Evaluation) | 72% | Win rate against competing OCR/document-AI systems across 600+ documents in 12+ languages |
| Blind Human Evaluation | 72% | Win rate vs. leading OCR/Document-AI systems (600+ multilingual documents) |
| Human evaluation | 72% win rate | Win rate in blind tests against competing OCR solutions |
| Human evaluation (blind test) | 72% | Win rate against competing OCR systems across 600+ documents in 12+ languages |
| Human Evaluation (Annotator Win Rate) | 72% | Average win rate over leading OCR systems in blind testing across 600+ documents in 12+ languages |
| Human Preference Evaluation | 72% | Win rate in blind human evaluations vs competing OCR solutions |
| Blind human evaluation | 72% | Win rate against competing OCR systems in independent annotator tests |
| Human Annotator Preference | 72% | Win rate in blind tests vs. leading OCR systems |
| Blind annotator evaluation | 72% | Independent annotators preferred OCR 4 over every leading OCR and document-AI system tested (600+ documents) |
| Blind human evaluation win rate | 72% | Against leading OCR and document-AI systems, 600+ documents |
| Human evaluation (win rate) | 72% | Independent annotators, 600+ documents |
| Annotator preference (blind evaluation) | 72% | Win rate across 600+ documents vs leading OCR and document-AI systems |
| Human evaluation win rate | 72% | Independent annotators preferred OCR 4 over every system tested |
| OlmOCRBench leaderboard | 85.20 | Tops the leaderboard |
Compare Mistral OCR 4
Compare Mistral OCR 4 with any other model
Build a comparison →All model comparisons →