TEST RECORD · ← all field reports

FIELD REPORTS/IMAGE/ISSUE #90

Answers questions about images and reads full pages of text

Liquid AI shrank a full vision model far enough that the official demo runs entirely inside a browser tab. Nothing installs and nothing uploads; the model itself downloads into the page and your images stay on your machine. That changes who gets to use a tool like this.

MODELLFM2.5-VL-3B
PUBLISHEDAugust 13, 2026
READ TIME3 min
TESTED BYNeural Expedition
CATEGORYIMAGE

Field notes

01What it does

You hand it an image and type what you want to know: what does this chart say, read this receipt, which items on this shelf are on sale. It answers in text, and it is unusually good at dense pages for its size: it can transcribe a full page while keeping the layout, so a two-column scan does not come back as one scrambled paragraph. Ask it to find something, "the red car", "the signature line", and it points to where that thing sits in the image. It follows questions in 16 languages, English and Chinese among them, and can compare several images in one conversation.

The unusual part is where it runs. The model fits in about 3 GB of memory, small enough for laptops and phones, which is why the demo can run it inside the browser itself instead of on a server. For anything private, receipts, medical letters, ID scans, that locality is the feature: the image never leaves your machine.

02How to try it

Open the WebGPU demo Space in Chrome or Edge and give it the nastiest document photo you have: a crumpled receipt, a two-column PDF screenshot, a whiteboard shot at an angle. Ask for the full text and watch whether the layout survives; that is where small vision models usually fall apart and where this one holds. For your own setup, the weights run through the usual local stacks including llama.cpp and MLX, with ready-made quantized versions for CPU-only machines.

03Caveat

It is built for quick, single questions, not long chains of reasoning: the developers themselves say reasoning-heavy jobs like turning a design mockup into working layout are outside its lane. Give it one clear task per prompt and it shines; ask it to think in steps and a bigger model will beat it. The license is Liquid's own LFM license, free to use for most readers, but larger companies should read the terms before shipping it in a product.

04What you can do with it

  • Pull clean text out of receipts, forms, and whiteboard photos without a cloud OCR service.
  • Ask what a chart or dashboard screenshot actually says instead of squinting at it.
  • Process documents that must not leave the building, on the machine they live on.
  • Add "find the thing I named" image search to an app without a vision API bill.
  • Read menus, signs, and labels across languages from a phone camera shot.

Try the demo

View model page

Read this issue on neuralexpedition.com →