Solutions / Document extraction / Demo
Invoice text. Structured results.
Run a sample invoice through an open model. Inspect structured fields, source evidence, and validation checks. This demo uses text samples. The managed Qwen3.8 27B API also accepts text, png/jpeg/webp images, and pdfs (scanned or digital, up to 10 pages).
See image and PDF API examples and limitsTry the live invoice example
Sign in to try it when the demo model is online. Each account gets 20 live requests for its lifetime. If the model is offline, see the ungated invoice example.
Try the demoWhat you’re seeing
- Structured output
- An open model fills a versioned invoice schema. It returns null for fields the document doesn’t contain.
- Checks in code
- Dates, currency, line math, subtotal, and total are verified in code, separately from the model.
- Source evidence
- Each value is matched to the text it came from. Values without a match are flagged for review.
- Latency on one GPU
- Timings come from the live request, streamed as the model generates.
What setup can cover
- Your documents, fields, and business rules instead of synthetic samples
- Connect image and PDF API input to your document workflow; this demo uses text samples
- Accuracy and review rates measured on a held-out set of your examples
- Throughput and cost per accepted document at your real volume
- Data handling, retention, and deployment location agreed before you share files
Synthetic invoices can show the approach, not your accuracy. Evaluation on your documents is agreed during setup.
Get setup help