Models & PricingModel Cards
December 12, 2025
GAPREMIERv25.12
ARVA OCR
Our OCR service powering our Document AI stack. Layout-aware extraction with tables, handwriting, and 150+ languages.
Speed
Performance
Context
100MB
Price
$1.00 / $2.50 /1K pages
Try it
import sagea
client = sagea.Client(api_key="your-api-key")
result = client.ocr.process(
model="arva-ocr",
document=open("invoice.pdf", "rb"),
output_format="markdown"
)
print(result.text)Base URL: https://api.sagea.space/v1 — Auth: Authorization: Bearer $SAGEA_API_KEY
Specifications
| Max File Size | 100 MB |
| Supported Formats | PDF, PNG, JPG, TIFF, WEBP |
| Languages | 150+ |
| Output Formats | Markdown, JSON, Plain Text |
| Accuracy | 99.2% on standard benchmarks |
Other Models
Capabilities
ARVA OCR is SAGEA's premier document understanding service, with native paragraph-level bounding box extraction, structural block labels, and block-level confidence scores.
- BBox extraction: paragraph-level boxes via
POST https://api.sagea.space/v1/ocr/process. - Structured annotations: tables, columns, headers, key-value pairs, form fields.
- Handwriting: cursive + print with high accuracy.
- Multilingual: 150+ languages, complex layouts, 100 MB files.
Benchmarks
| Metric | Value |
|---|---|
| Accuracy (standard sets) | 99.2% |
| Languages | 150+ |
| Max file | 100 MB |
| Outputs | Markdown, JSON, text + bboxes |
Use cases
- Invoice / receipt / form extraction
- Document QnA pipelines with SAGE chat models
- Batch digitization via
POST https://api.sagea.space/v1/batch
Limitations
- Scanned handwriting below 150 DPI degrades accuracy — prefer 300 DPI.
- Complex docs billed at $2.50 / 1,000 pages. See Choosing a Model.
Next steps
- Document processing quickstart
- OCR cookbooks: invoices, tables, handwriting, batch
- API Reference