Can Mistral AI process a document with mixed digital and scanned pages?
Mistral AI processed the scanned page in a mixed digital and scanned document rather than skipping it. The page appeared in document order, with tables in the same style as the digital pages.
1 of 1 test case passed
Every test case in this scenario has a result.
The test case
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.
Mixed pages — scanned page not skippedPassEvidence
# Service worksheet | Job reference | KAL-ESR-4471 | | --- | --- |
Markdown output: the service report, including the scanned page
Copied from Output file (Markdown), lines 36–40
Open the Markdown file ↗| Engineer | R. Tarbert | | Date | 29 January 2026 | # CHECKS AND READINGS | Flow cell condition on arrival | Fouled, inner face | | --- | --- | | Flow cell pressure after service | 2.4 bar | | Zero drift over 30 minutes | 0.003 mg/l | | Span recovery against standard | 99.4 per cent | | Lamp hours at service | 8,140 | | Result | Pass | # Calibration certificate SR-4471/2 Analyser left in service on channel 2. Optical window to be inspected at the next visit. # TIME ON SITE | Arrived | 29 January 2026, 09:15 | | --- | --- | | Left | 29 January 2026, 13:40 | | Chargeable time | 4 hours 25 minutes | | Next service due | January 2027 | # RECEIVED BY | Name | M. Cairney | | --- | --- | | Position | Duty operator, Vardenne Quay | | Date | 29 January 2026 |
Copied from Output file (Markdown), lines 42–72
| ✓ | Expected: The scanned page’s text appears in the output like the digital pages’ text, rather than being silently skipped.Found: The scanned page appeared in order, with its text and tables rendered like the digital pages. The engineer’s signature remained an image reference. |
The scanned page had to appear alongside the digital pages. It was not skipped, and its tables used the same style. The source PDF is available, and the run card identifies it as the document sent, allowing the output to be compared with the source.
Heading-level drift. In the output, the scanned page’s checks and readings, time on site, and received by labels become level-1 headings, one level higher than page 1’s section headings.
Plain job reference. In the output, the scanned page’s job reference is plain text, while page 1’s is bold.
Ungraded first page. On page 1, where the source has a horizontal rule followed by the CLIENT label, the output emits a mangled heading ("# ---**CLIENT**"). Page 1 is not graded, so this does not affect the result.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | Converting a complex PDF into clean Markdown with a hosted API → | 12 scenarios · 13 tools |
| Capability | Scanned Document OCR → | |
| Scenario | A document mixing digital and scanned pages → | |
| Tool | Mistral AI → |
Global scenario definition → · Global capability definition → · Mistral AI product page →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.