Unlimited OCR 2
Specifications
- Input
- Output
- Context window
- 83K tokens
- Released
- Nov 2026
Performance
- Speed
- 445 t/s
- TTFT
- 530 ms
- Latency
- 150 ms
- Intelligence
- —
Pricing
- Input
- €0.10 per 1M tokens
- Output
- €0.20 per 1M tokens
About this model
Baidu's Unlimited-OCR-2 (3B, MIT), an OCR model for scanned pages and photos of documents. Send up to 20 page images per request on POST /v1/chat/completions and get each page's text with its layout blocks, one <PAGE> marker per page. Billed per token; a typical page costs about EUR 0.30 per 1,000 pages.
Technical specifications
- Capabilities
- Input modalities
- Output modalities
- Reasoning
- No
See also
Add Model to Comparison
Search for a model to add