Azure Document Intelligence vs Qwen-VL
Pick Azure Document Intelligence for invoices, receipts, IDs and other fixed-format documents, especially if you already run on Azure or need disconnected containers on-prem. Pick Qwen-VL when you need OCR plus reasoning or VQA in one model, or want Apache-2.0 weights you can self-host. The main split is a managed, priced-per-page extraction service versus a free general vision model that can loop on bad scans.
Azure Document Intelligenceazure/document-intelligence | Qwen-VLqwen/qwen-vl | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 60 credits / page | 2 credits / page |
| Per 1,000 pages | $60.00 | $2.00 |
| Typical speed | 1.5 s / page | 2.0 s / page |
| Best for | Enterprise forms and IDs | Flexible extraction with prompts |
| Inputs | PDF, images | PDF, images |
| Languages | 300+ languages | 30+ languages |
| Max pages per request | 2,000 | 200 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | No | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | ✓ Yes | No |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose Azure Document Intelligence if
- It ships prebuilt models for invoices, receipts, IDs, W-2 and contracts, so you do not have to build extraction logic yourself.
- Read OCR costs $1.50 per 1,000 pages and drops to $0.60 above 1M pages.
- Disconnected containers let you run it on-prem, and it fits teams already on Azure.
Choose Qwen-VL if
- The weights are Apache-2.0 and come in sizes from 2B to 235B-A22B MoE, so you can fit the model to your hardware.
- It handles OCR in 32 languages and also does reasoning and VQA, so one model covers photos, screenshots and documents.
- It outputs layout positions and a Qwen HTML document format, as well as plain text, JSON, Markdown and bounding boxes.
Key differences
| Azure Document Intelligence | Qwen-VL | |
|---|---|---|
| Licence | Proprietary | Apache-2.0 (Qwen3-VL repository) |
| How you run it | Managed Azure service, or containers including disconnected on-prem use | Self-host the open weights, or use Alibaba Cloud's hosted API |
| Price | Read $1.50 per 1,000 pages ($0.60 above 1M); Layout and prebuilt $10 per 1,000; custom extraction $30 per 1,000; 500 free pages per month | Free to self-host, you pay for compute. Hosted pricing not verified |
| Output formats | JSON, Markdown, plain text, bounding boxes, searchable PDF | Plain text, HTML, bounding boxes, JSON, Markdown |
| Benchmark | 82.7% on RD-TableBench table accuracy (independent) | 64.6 on olmOCR-bench for Qwen 3 VL 8B (independent), below dedicated OCR models |
| Hardware | Not published | FP8 checkpoints need an H100-class GPU and CUDA 12+; the 235B example uses 8-way tensor parallel on vLLM |
| Known failure | Stacked models multiply cost: Prebuilt plus Layout plus high-res add-on is $26 per 1,000 pages. The free tier only processes the first two pages of each document. | Can loop and repeat the same text, worse with greedy decoding and on low-quality scans. One report saw loops on 10-37% of OCR outputs after a transformers upgrade, versus 2-3% before. |
Sources
- Azure pricing page
- Reducto RD-TableBench
- Qwen3-VL GitHub
- Issue #1611 infinite repetition
- olmo-eval issue #441
- Chandra README benchmark table
What each one is
Azure Document Intelligence
Azure layout model with tables, selection marks and prebuilt invoice and ID readers.
Azure Document Intelligence detailsQwen-VL
Alibaba's open-weight vision language models (Qwen2.5-VL, Qwen3-VL). Strong general OCR and grounding.
Qwen-VL detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "azure/document-intelligence","fallbacks": ["qwen/qwen-vl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Azure Document Intelligence alternativesQwen-VL alternativesBest OCR models in 2026