AI & Enterprise
China AI startup Z.ai open-sources GLM-OCR document understanding model
Chinese AI startup Z.ai has open-sourced GLM-OCR, a multimodal optical character recognition model tailored for document understanding. The company says the 900 million-parameter model analyses layouts and recognises text in two stages, aiming to improve accuracy on complex documents. Z.ai reported a 94.62 score on OmniDocBench V1.5 and said the model can run locally, processing PDFs and images quickly and exporting results as HTML or JSON.
