$5 free credits when you sign up Claim now
Marlin 2b now available Test it!
MiniMax H3 now available! Test it!
Browse all Models
Browse all
Browse all
Browse all
  • Image-to-Text

LightOnOCR

LightOn's open-weight document OCR model. Turns scans, photos and screenshots into clean Markdown, with tables as HTML and equations as LaTeX. Reads print and handwriting in many languages.

  • Free tier available
  • No subscriptions
  • No credit card required

Sample Outputs

Real outputs generated with LightOnOCR on deAPI.

Capabilities

What you can do with LightOnOCR.

Image-to-Text

Extract text from images with high accuracy. Ideal for document processing, receipt scanning, and automated data entry.

Technical Specifications

Resolution

128–4096 x 127–4096 px

How to use in deAPI

Integrate LightOnOCR into your app with a single API call.

POST https://api.deapi.ai/api/v2/images/ocr

curl · deAPI images/ocr

curl -X 'POST' \
  'https://api.deapi.ai/api/v2/images/ocr' \
  -H 'accept: application/json' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -F 'model=LightOnOcr_3_4B_Q8' \
  -F 'image=@/path/to/file'{
    "model": "LightOnOcr_3_4B_Q8",
    "image": "@/path/to/image.png"
}

Tip: The API returns a request_id. Use webhooks (recommended) or poll GET /jobs/{request_id} for results.

Need an API key?

Sign up for free and get $5 in credits to start.

Get $5 credits

Frequently Asked Questions

LightOn's open-weight document OCR model. Turns scans, photos and screenshots into clean Markdown, with tables as HTML and equations as LaTeX. Reads print and handwriting in many languages.
Yes. On a handwritten letter from 1864 it transcribed every word correctly. Note that it produces a readable transcript, not a diplomatic one: it normalizes punctuation, capitalization and words split across line breaks. If you need the exact spelling and line breaks of a manuscript, review the output against the original.
It reads Latin-script languages with their accents and special characters intact. In our tests Polish, French, German, Spanish and English came through without a single wrong character, and you do not need to tell the model which language the page is in.
JPG, JPEG, PNG, GIF, BMP or WebP files up to 10 MB, from 128 to 4096 pixels per side. For documents, around 2048 pixels on the longer side (an A4 page at about 175 DPI) is a good balance of accuracy and cost; that is also the resolution LightOn recommends for the model.
deAPI is pay-per-use with no subscriptions. You pay per image, based only on its pixel count: about $0.0074 for a 1240×663 screenshot, $0.0093 for a 1024×1024 image, $0.0188 for an A4 page at 150 DPI (1240×1754) and $0.0254 for an A4 page at 2048 px. The amount of text on the page does not change the price. Quote any image in advance with POST /api/v2/images/ocr/price, or see https://deapi.ai/pricing.
Yes. LightOnOCR-3-4B is released by LightOn under the Apache 2.0 license. On deAPI you use it through the API without hosting anything yourself. LightOn reports scores of 86.3 on olmOCR-Bench and 75.1 on ParseBench for this model.
Yes. LightOnOCR-3-4B is available via the deAPI REST API. Submit an image with `"model": "LightOnOcr_3_4B_Q8"`, receive a request ID, and get the result by webhook or by polling. See the API documentation at https://docs.deapi.ai for code examples in Python, Node.js, and cURL.
  • No subscription
  • No credit card required
  • Pay only for what you use

Try LightOnOCR — $5 for free

Get $5 in free credits and start generating with LightOnOCR.