OCR models / OpenDataLab (Shanghai AI Lab)
MinerU2.5 Pro (2605)
OpenDataLab (Shanghai AI Lab) · released 2026-05-20 · data checked 2026-11-06
MinerU2.5 Pro (2605) is a 1.2B-parameter OCR and document-parsing model from OpenDataLab (Shanghai AI Lab) (CN), released 2026-05-20 under the Apache-2.0 license. It scores 95.72 on OmniDocBench v1.6 (#8 of 33). It needs about 3.3 GB of memory at 16-bit.
PDF to Markdown with the MinerU toolkit. Chinese and English.
OmniDocBench v1.6
95.72
#8 of 33 models
olmOCR-Bench
—
Not reported
Parameters
1.2B
OCR specialist
Memory to run
~3.3 GB
16-bit weights plus about 1 GB
Benchmarks
OmniDocBench v1.6 Lab-reported95.72
The leaderboard lists the earlier 2604 release at 95.75. Best: OXR 1.0 97.03
The tick marks the best score on each benchmark. See how to read the ranking.
Specifications
- Lab
- OpenDataLab (Shanghai AI Lab) (CN)
- Type
- OCR / document parsing specialist
- Total parameters
- 1.2B
- License
- Apache-2.0
- Memory to run
- ~3.3 GB at 16-bit
- Released
- 2026-05-20
Closest alternatives
- GLM-OCR (1.3B, 95.22 on OmniDocBench, MIT)
- PaddleOCR-VL 1.6 (960M, 96.34 on OmniDocBench, Apache-2.0)
- OvisOCR2 (850M, 96.47 on OmniDocBench, Apache-2.0)
- PaddleOCR-VL 1.5 (960M, 94.93 on OmniDocBench, Apache-2.0)
Sources
- https://huggingface.co/opendatalab/MinerU2.5-Pro-2605-1.2B
- https://github.com/opendatalab/OmniDocBench
Something wrong here? Report it.