New OCR Models: LightOnOCR-2 and GLM-OCR Improve Accuracy

A LocalLLaMA user has expressed appreciation for the new OCR models LightOnOCR-2 and GLM-OCR, reporting a significant improvement over previous versions available in late 2025.

Model Details

• LightOnOCR-2-1B: Documentation is available at https://docs.voxel51.com/plugins/plugins_ecosystem/lightonocr_2.html. A quickstart notebook is available at https://github.com/harpreetsahota204/LightOnOCR-2/blob/main/lightonocr2_fiftyone_example.ipynb.
• GLM-OCR: Documentation is available at https://docs.voxel51.com/plugins/plugins_ecosystem/glm_ocr.html. A quickstart notebook is available at https://github.com/harpreetsahota204/glm_ocr/blob/main/glm_ocr_fiftyone_example.ipynb.

According to the user, GLM-OCR stands out for its speed and ability to provide reliable structured output. OCR (Optical Character Recognition) is a fundamental technology for converting images of text, such as scanned documents or photographs, into textual data usable by computers. Improvements in this field have a direct impact on process automation and information accessibility.