cd /news/large-language-models/you-sent-the-delivery-note-photos-at… · home › topics › large-language-models › article
[ARTICLE · art-172413] src=dev.to ↗ pub= topic=large-language-models verified=true sentiment=↓ negative

You Sent the Delivery Note Photos at Low Detail and the Model Read a Thumbnail

Serguey Asael Shinder reports that a delivery-note extraction model that tested well on office scans failed in production because the image calls used OpenAI's "detail": "low" setting, which downsamples photos to a fixed 512x512 thumbnail and strips small print and handwriting. The model filled every field with plausible but wrong values — invented batch codes and misread quantities — without flagging the images as unclear. The fix was switching documents to high detail, cropping photos to the page, building a test set from real driver photos, and requiring unreadable fields to return empty rather than guessed.

recommendation

Check what resolution your model actually receives, use high detail for documents, crop to the region of interest, and make unreadable fields return empty instead of guessed.

more in playbook: Verifying model output
by read2 min views5 publishedNov 19, 2026

The drivers photographed

every signed delivery note.

A model read each photo

and pulled out

the order number,

the quantities,

and any note about damage.

In testing

it was excellent.

In the first week live

the cost was lower than planned.

Then the complaints came.

Quantities wrong.

Batch codes invented.

A handwritten "2 bags split"

read as "2 bags".

Same model.

Same prompt.

Nothing in the response

said the photo was unclear.

Every field came back filled in,

with the same confidence

as before.

Here is the mechanism.

The image call

had one extra field:

"detail": "low" Someone had added it

to save money.

On OpenAI's API,

low detail

does not mean

"look a little less carefully".

It means the model receives

one small copy of the image,

512 pixels by 512.

A fixed, small number of tokens.

Whatever the size

of the photo you sent.

A twelve megapixel phone shot

of an A4 page

becomes a thumbnail.

The printed headings survive.

The order number,

in large type,

survives.

The small print does not.

Neither does handwriting.

So the model

did what models do

with a blurred field.

It produced

something plausible.

Testing had used

scans from the office,

cropped and flat.

The drivers sent

whole cab dashboards

with a note in the corner.

Other providers

shrink large images too,

just at different sizes.

So the questions are always the same.

What resolution

does the model actually see?

And is the part you care about

still readable at that size?

What fixed it:

Detail set to high

for documents. Low kept

for "is this a photo of a pallet". Photos cropped

to the page

before sending.

A test set

made of real driver photos,

not office scans.

And a rule

that any field

the model could not read clearly

comes back empty,

not guessed.

The model did not misread

your delivery note.

It never saw it.

You sent it a postage stamp

and asked for the small print.

– Serguey Asael Shinder

── more in #large-language-models 4 stories · sorted by recency
suno.com · · #large-language-models
Suno MCP
── more on @openai 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/you-sent-the-deliver…] indexed:0 read:2min 2026-11-19 · —