OCR models / Baidu (PaddlePaddle)
PaddleOCR-VL 1.6
Baidu (PaddlePaddle) · released 2026-05-27 · data checked 2026-11-06
PaddleOCR-VL 1.6 is a 960M-parameter OCR and document-parsing model from Baidu (PaddlePaddle) (CN), released 2026-05-27 under the Apache-2.0 license. It scores 96.34 on OmniDocBench v1.6 (#7 of 33). It needs about 2.9 GB of memory at 16-bit.
Covers 100+ languages and runs in about 2 GB of GPU memory. Pairs the 0.9B model with the PP-DocLayout layout detector.
OmniDocBench v1.6
96.34
#7 of 33 models
olmOCR-Bench
—
Not reported
Parameters
960M
OCR specialist
Memory to run
~2.9 GB
16-bit weights plus about 1 GB
Specifications
- Lab
- Baidu (PaddlePaddle) (CN)
- Type
- OCR / document parsing specialist
- Total parameters
- 960M
- License
- Apache-2.0
- Memory to run
- ~2.9 GB at 16-bit
- Released
- 2026-05-27
Closest alternatives
- OvisOCR2 (850M, 96.47 on OmniDocBench, Apache-2.0)
- OvisOCR3 (850M, 96.74 on OmniDocBench, Apache-2.0)
- Xiaomi-OCR-0 (870M, 96.83 on OmniDocBench, Apache-2.0)
- TeleOCR (1.4B, 96.91 on OmniDocBench, Apache-2.0)
Sources
Something wrong here? Report it.