OCR models / Z.ai
GLM-OCR
Z.ai · released 2026-01-30 · data checked 2026-11-06
GLM-OCR is a 1.3B-parameter OCR and document-parsing model from Z.ai (CN), released 2026-01-30 under the MIT license. It scores 95.22 on OmniDocBench v1.6 (#9 of 33) and 75.2 on olmOCR-Bench (#19 of 24). It needs about 3.7 GB of memory at 16-bit.
Uses the PP-DocLayoutV3 layout model (Apache 2.0) in its pipeline.
OmniDocBench v1.6
95.22
#9 of 33 models
olmOCR-Bench
75.2
#19 of 24 models
Parameters
1.3B
OCR specialist
Memory to run
~3.7 GB
16-bit weights plus about 1 GB
Benchmarks
olmOCR-Bench Leaderboard75.2
Run through the Z.ai API; headers and footers excluded. Best: OXR 1.0 87.8
The tick marks the best score on each benchmark. See how to read the ranking.
Specifications
- Lab
- Z.ai (CN)
- Type
- OCR / document parsing specialist
- Total parameters
- 1.3B
- License
- MIT
- Memory to run
- ~3.7 GB at 16-bit
- Released
- 2026-01-30
Closest alternatives
- PaddleOCR-VL 1.5 (960M, 94.93 on OmniDocBench, Apache-2.0)
- HunyuanOCR 1.5 (1.1B, 94.74 on OmniDocBench, Tencent Hunyuan Community License)
- MinerU2.5 Pro (2605) (1.2B, 95.72 on OmniDocBench, Apache-2.0)
- MiniCPM-V 4.7 35B-A3B (35.2B, 94.24 on OmniDocBench, Apache-2.0)
Sources
Something wrong here? Report it.