Models
qomplement provides OCR and Fill models. Choose the right model based on your accuracy needs and use case.
GET /v1/models
Available Models
qomplement-OCR-v1
The default model. Fast, cost-effective, and accurate for most documents.
| Property | Value |
|---|---|
| Model ID | qomplement-OCR-v1 |
| Type | Standard OCR + AI structuring with advanced layout analysis |
| Best for | Digital PDFs, clean scans, typed documents, invoices, contracts, and structured business documents |
| Speed | ~2-4 seconds per page (varies by document complexity and image resolution) |
| Token cost | 1x (base rate) — most cost-effective option for high-volume processing |
Capabilities:
| Feature | Status |
|---|---|
| Text extraction | Fully supported with advanced layout preservation and reading order detection |
| Table detection | Fully supported including complex multi-row headers, merged cells, and nested table structures |
| Key-value pairs | Fully supported with semantic understanding and contextual relationship detection |
| Schema-guided extraction | Fully supported with custom field definitions, validation rules, and confidence scoring |
| Multi-page documents | Fully supported up to 200 pages per request with automatic page range processing |
| Handwriting recognition | Limited support — works best for printed text; may struggle with cursive or poor handwriting |
| Low-quality scans | Limited support — requires minimum 150 DPI; degraded or faded documents may need OCR-XL model |
qomplement-OCR-XL-v1
High-precision model for complex or degraded documents. Uses enhanced multi-pass processing for maximum accuracy.
| Property | Value |
|---|---|
| Model ID | qomplement-OCR-XL-v1 |
| Type | Enhanced OCR + multi-pass AI structuring with deep learning handwriting and degraded document processing |
| Best for | Handwritten documents, low-quality scans, faded or damaged documents, complex medical/legal forms, historical archives |
| Speed | ~5-10 seconds per page (longer for heavily degraded or handwritten content with multiple verification passes) |
| Token cost | 2x (double base rate) — premium pricing for maximum accuracy and specialized document handling capabilities |
Capabilities:
| Feature | Status |
|---|---|
| Text extraction | Fully supported with enhanced character recognition for degraded, blurred, or faded text across all document types |
| Table detection | Fully supported including irregular tables, hand-drawn borders, and documents with mixed handwritten/printed content |
| Key-value pairs | Fully supported with advanced contextual analysis and relationship detection even in unstructured layouts |
| Schema-guided extraction | Fully supported with intelligent field matching, fuzzy text recognition, and adaptive confidence thresholds |
| Multi-page documents | Fully supported up to 200 pages per request with cross-page table continuation and section detection |
| Handwriting recognition | Fully supported including cursive, print, mixed styles, and multiple languages with deep learning verification |
| Low-quality scans | Fully supported down to 75 DPI with noise reduction, contrast enhancement, and adaptive preprocessing |
qomplement-fill-v1
AI-powered form filling model. Detects PDF form fields and intelligently maps data from source documents, natural language instructions, or explicit mappings.
| Property | Value |
|---|---|
| Model ID | qomplement-fill-v1 |
| Type | AI-powered form field detection + intelligent semantic mapping with cross-document data extraction |
| Best for | PDF form filling, document-to-form automation, government forms, application processing, medical intake forms, legal document generation |
| Speed | ~3-8 seconds per form (varies by number of fields, source document complexity, and mapping method used) |
| Token cost | 1x (base rate) — cost-effective for both single form fills and high-volume batch processing workflows |
Capabilities:
| Feature | Status |
|---|---|
| Form field detection | Fully supported with automatic identification of text fields, checkboxes, radio buttons, dropdowns, and signature fields |
| Source document extraction | Fully supported with intelligent data extraction from multiple source documents and automatic field-to-value matching |
| Natural language instructions | Fully supported with plain English commands like "Fill applicant name as John Smith" for simple and complex filling logic |
| Explicit JSON mappings | Fully supported with direct field-to-value assignments using JSON format for programmatic control and batch processing |
| Cross-language field matching | Fully supported with semantic understanding across English, Spanish, French, German, and 20+ other languages |
| Confidence scoring | Fully supported with per-field confidence metrics (0-100%) to identify uncertain mappings requiring manual review |
| Form flattening | Fully supported with option to flatten filled PDFs into non-editable documents for final distribution and archival |
Three input methods:
| Method | Description |
|---|---|
| Source documents | Upload one or more source documents (PDFs, images, Word files) and the AI automatically extracts relevant data and intelligently maps it to matching form fields using semantic understanding and field name analysis |
| Instructions | Describe what to fill using plain language commands (e.g., "Fill client name as 'Acme Corp' and set date to today") — ideal for simple forms or when you need to override specific fields with custom values |
| Explicit mappings | Provide exact field-to-value mappings using JSON format like {"field_name": "value"} — gives you complete programmatic control for batch processing, API integrations, and database-driven form filling workflows |
Usage
Specify the OCR model in extraction or fill endpoints using the model parameter:
curl -X POST https://developer-api.qomplement.com/v1/extract \
-H "Authorization: Bearer sd_your_api_key" \
-F "files=@document.pdf" \
-F "model=qomplement-OCR-XL-v1"
If omitted, the API defaults to qomplement-OCR-v1. The fill endpoint automatically uses qomplement-fill-v1 for field mapping.
Model Versions
Models are versioned to ensure consistent behavior. The current versions are:
| Alias | Points to | Released |
|---|---|---|
qomplement-OCR-v1 | qomplement-OCR-v1-2025-02 — Latest standard OCR model with improved table detection and layout analysis | February 2025 |
qomplement-OCR-XL-v1 | qomplement-OCR-XL-v1-2025-02 — Latest high-precision model with enhanced handwriting recognition and degraded document processing | February 2025 |
qomplement-fill-v1 | qomplement-fill-v1-2025-02 — Latest form filling model with improved semantic matching and multi-language field detection | February 2025 |
When new versions are released, the alias automatically points to the latest version. Pin a specific version by using the full version ID (e.g., qomplement-OCR-v1-2025-02) if you need deterministic behavior.
Rate Limits
Rate limits ensure fair access to the API. Your tier determines your limits.
| Tier | Monthly Tokens | Requests / min | Max Pages / request | Max File Size |
|---|---|---|---|---|
| Free | 100K tokens — suitable for testing and small projects with approximately 100-200 pages per month | 10 requests/min — ideal for development and prototyping workflows | 10 pages maximum per single API request | 10 MB per file — sufficient for most standard documents |
| Starter | 500K tokens — supports small businesses processing 500-1,000 pages monthly with room for growth | 30 requests/min — enables moderate automation and scheduled batch processing | 50 pages maximum per single API request | 25 MB per file — handles larger documents and high-res scans |
| Growth | 2.5M tokens — designed for growing teams processing 2,500-5,000 pages per month at scale | 60 requests/min — supports high-volume automation, webhooks, and real-time integrations | 100 pages maximum per single API request | 50 MB per file — accommodates complex multi-page reports |
| Scale | 10M tokens — enterprise-grade capacity for 10,000-20,000 pages monthly with dedicated support | 120 requests/min — handles enterprise workloads with parallel processing and mission-critical SLAs | 200 pages maximum per single API request | 100 MB per file — processes large archives and consolidated documents |
Rate limit headers are included in every response:
X-RateLimit-Limit: 30
X-RateLimit-Remaining: 29
X-RateLimit-Reset: 1709078400
If you exceed your rate limit, the API returns 429 Too Many Requests. See Rate Limiting for details.
Pricing
Tokens are consumed per page processed. The XL model uses 2x tokens per page.
| Model | Tokens per page/form | Approximate cost |
|---|---|---|
qomplement-OCR-v1 | ~500-1,500 tokens per page depending on document complexity, table density, and content structure | Varies by pricing tier — most economical option for high-volume batch processing and standard document extraction |
qomplement-OCR-XL-v1 | ~1,000-3,000 tokens per page with enhanced processing for handwriting and degraded documents | 2x standard rate — premium pricing justified by superior accuracy for specialized documents requiring maximum precision |
qomplement-fill-v1 | ~500-2,000 tokens per form based on number of fields, source document complexity, and mapping method used | Varies by pricing tier — competitive rates for both single-form fills and large-scale automated form processing pipelines |
Purchase tokens on the Developer Portal:
| Package | Tokens | Price | Per 1M tokens |
|---|---|---|---|
| Starter | 500K tokens — ideal for small businesses and development teams starting with document automation | $15 per month — affordable entry point with no long-term commitment required | $30 per million tokens — competitive rate for testing and small-scale deployments |
| Growth | 2.5M tokens — designed for growing businesses scaling their document processing workflows | $60 per month — best value tier with 20% savings compared to Starter pricing | $24 per million tokens — optimized pricing for medium-volume production workloads |
| Scale | 10M tokens — enterprise-grade capacity with priority support and dedicated account management | $200 per month — maximum volume discount with premium SLA guarantees and custom integrations | $20 per million tokens — lowest per-token cost for high-volume enterprise customers |
Choosing a Model
| Scenario | Recommended Model |
|---|---|
| Digital PDFs with typed text | qomplement-OCR-v1 — optimized for born-digital documents, invoices, contracts, and business correspondence with fast processing and excellent accuracy for machine-generated text |
| Clean scans (300+ DPI) | qomplement-OCR-v1 — handles high-quality scanned documents efficiently with precise text extraction, table detection, and layout preservation at the most cost-effective rate |
| Handwritten forms | qomplement-OCR-XL-v1 — specialized deep learning models trained on cursive, print, and mixed handwriting styles with multiple verification passes for maximum accuracy |
| Faded or damaged documents | qomplement-OCR-XL-v1 — advanced preprocessing with noise reduction, contrast enhancement, and adaptive thresholding to recover text from degraded historical or archived documents |
| Mixed typed + handwritten | qomplement-OCR-XL-v1 — intelligently segments and processes different text types using appropriate recognition engines for optimal results across content variations |
| High-volume batch processing | qomplement-OCR-v1 — fastest processing speed and lowest token cost makes it ideal for processing thousands of pages daily in automated production workflows |
| Legal/medical documents requiring accuracy | qomplement-OCR-XL-v1 — highest precision model with multi-pass verification recommended for mission-critical documents where errors can have serious compliance or liability consequences |
| PDF form filling from source documents | qomplement-fill-v1 — automatically extracts data from source documents and intelligently maps to PDF form fields using semantic field matching and contextual understanding |
| Form filling with natural language instructions | qomplement-fill-v1 — accepts plain English commands to fill forms programmatically, ideal for simple forms or when you need human-readable filling logic in your codebase |
tip
Start with qomplement-OCR-v1 for most use cases. Switch to qomplement-OCR-XL-v1 only if you see accuracy issues with specific document types. You can use the Playground to compare results side by side.