Sarvam Vision 2.1
Vision 2.1 adds document capabilities for complex tables, form fields and Indian-language handwriting. The current Document AI overview identifies it as the model behind its 23-language service.
The document system includes layout components
Sarvam combines the vision-language model with a semantic layout parser and a pointer-based reading-order network. The release describes supervised post-training followed by reinforcement learning with verifiable rewards.Sarvam Vision 2.1 release
Practical use
Practical guidance. Evaluate fields, reading order and table structure separately. A correct sequence of words can still produce an unusable invoice if a total is attached to the wrong row or form field.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Model identifiers | sarvam-vision |
Related models
- Sarvam Translate V1 API
Sarvam AI ยท Sarvam Translate
The Sarvam Translate API supports English and all 22 scheduled Indian languages.
Launch date not confirmedAPI