← all releases
Baidu·Jan 29, 2026·v1.5·open source
PaddleOCR-VL-1.5
PaddleOCR-VL-1.5 is an advanced model in the PaddleOCR-VL series, achieving a state-of-the-art accuracy of 94.5% on OmniDocBench v1.5. It introduces robust document parsing capabilities, including support for irregular-shaped localization and text spotting tasks.
OCR
overview
PaddleOCR-VL-1.5 is an advanced model in the PaddleOCR-VL series, achieving a state-of-the-art accuracy of 94.5% on OmniDocBench v1.5. It introduces robust document parsing capabilities, including support for irregular-shaped localization and text spotting tasks.
key features
- 0194.5% accuracy on OmniDocBench v1.5
- 02Supports irregular-shaped localization for skewed and warped documents
- 03Includes text spotting and seal recognition tasks
- 04Compact 0.9B parameter size for high efficiency
- 05Enhanced performance in real-world scenarios like scanning and illumination changes
related productsbrowse all →