Unlimited OCR Models by Baidu AI
Baidu AI releases · 9 in the last 12 months · this family litChangelog →
1 model2026
Last refreshed 2026-10-27. Next refresh: weekly.
Capabilities
VisionAll models
MultimodalAll models
About
Baidu's open document-parsing (OCR) models that parse documents of unbounded length in one pass using sliding-window attention (SR-SWA in version 2).
Decision facts
- Best fit
- ocrvision and multimodal work
- Capability starting point
- Unlimited OCR 2 with multimodal inputs
- Lowest tracked input
- Not tracked
- Closest related family
- ERNIE 4.5
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
1 in view
Unlimited OCR 2Current
Use when the workload needs ocr, 3.3B parameters, and multimodal inputs.
2026-10ocr3.3B parametersmultimodal inputs
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Unlimited OCR 2 | Use when the workload needs ocr, 3.3B parameters, and multimodal inputs. | 2026-10 | ocr3.3B parametersmultimodal inputs | Current |
Release Timeline
1 release group2026-10
1 current
Unlimited OCR 2
Currentocr3.3B parametersmultimodal inputs
Specifications(1 models)
| Model | Released | Parameters | Vision | Multimodal |
|---|---|---|---|---|
| Unlimited OCR 2 | 2026-10 | 3.3B | Yes | Yes |
