AI on Demandpowered by LightOn

LightOnOCR-2-1B: 9 times smaller than comparable OCR models — and up to 3 times faster. An end-to-end model from the French company LightOn AI, operated by stepping stone on Swiss infrastructure.

LightOnOCR-2-1B is an end-to-end optical character recognition model developed by the French company LightOn AI. With just 1 billion parameters, it is around 9 times smaller than comparable models — yet faster: up to 3 times faster processing speed whilst delivering state-of-the-art results.

The model reliably recognises text in complex layouts, scans, academic documents, tables and forms. stepping stone runs LightOnOCR entirely on Swiss infrastructure. Access is via an OpenAI-compatible API, which can be integrated directly into existing workflows. Your documents remain in Switzerland.

Companies and public authorities wishing to digitise large volumes of documents efficiently — quickly, accurately and without transferring data to US providers. Particularly suitable where speed and cost-effectiveness are crucial.

Typical applications: bulk processing of invoices, contracts and forms; digitisation of archives and legacy records; extraction from multi-column layouts and academic documents; OCR integration into automated document pipelines.

Open source (Apache 2.0). European model. Swiss data centres. No data stored with US providers.

Compact and efficient: LightOnOCR processes up to 493,000 pages per day on a single GPU — for less than CHF 0.01 per 1,000 pages. No external OCR toolkit required; everything is end-to-end. Personalised advice and operation provided by stepping stone in Bern.

Your Consultant

Scope of services

On-demand document recognition

Access to LightOnOCR for fast, accurate text recognition. Excellent for tables, forms, multi-column layouts, academic documents and scans. Up to three times faster than comparable models.

GPU performance on demand

Scalable computing power for individual documents or entire archives. Particularly cost-effective thanks to the compact model — you pay as you go.

Managed service

Deployment, monitoring, maintenance and support on Swiss infrastructure, with personalised advice. stepping stone takes care of the day-to-day running so that you can focus on the benefits.

Areas of application

Batch processing

LightOnOCR is designed for high throughput — up to 493,000 pages per day on a single GPU.

Companies use this to process large volumes of invoices, contracts and forms efficiently and cost-effectively. Even extensive archive collections can be digitised cost-effectively for less than CHF 0.01 per 1,000 pages. 

Document pipelines

LightOnOCR integrates seamlessly into automated document processing workflows as a compact module.

Thanks to its OpenAI-compatible API, it can be integrated directly into existing workflows — without the need for an external OCR toolkit. It processes complex layouts, tables, forms and academic documents end-to-end on Swiss infrastructure.

Benchmark

The benchmark processes 50 CVs (100 pages total). Step-by-step instructions and the required Python script can be downloaded from GitHub.

If necessary, higher concurrency and page limits can be set.

Call

# Set your personal key:
STONEY_KEY=sk-...

# Make key visible for bench script:
export OPENAI_API_KEY=$STONEY_KEY

# Start the benchmark
python cv_bench_endpoint.py \
 --endpoint llm.stoney-cloud.com/v1/chat/completions \
 --data cv_bench_data \
 --model "lightonai/LightOnOCR-2-1B" \
 --api-key $STONEY_KEY \
 --concurrency 1 \
 --limit 100

Result

concurrency   : 1
requested     : 50
ok            : 50
failed        : 0
duration_s    : 93.958
pages_s       : 0.532
pages_min     : 31.9
out_tok_s     : 419.4
latency_p50_s : 1.63
latency_p99_s : 10.016

 

Legend

  • concurrency: How many requests the model processes simultaneously.
  • requested: How many requests were sent.
  • ok: Number of accepted requests (in this case, CVs).
  • failed: Number of rejected requests.
  • duration_s: The duration of the benchmark run.
  • pages_s: The average number of pages that can be processed per second.
  • pages_min: The average number of pages that can be processed per minute.
  • out_tok_s: The number of tokens generated per second.
  • latency_p50_s: The average response time in seconds.
  • latency_p99_s: The response time required in the “worst case” scenario, in seconds.

Price

ModelContext lengthInput/MTokOutput/MTok
LightOnOCR-2-1B16k0.02000.0600
All prices are in CHF/MTok, excluding VAT.