Understand this story
On October 26, 2026, researchers from Tsinghua University and Zhipu AI published the paper "Can Visual Generation Expand What Unified Multimodal Models Know?" on arXiv. The corresponding author is Zhipu Chief Scientist Tang Jie. The paper introduces GLM-Omni, a previously undisclosed internal Zhipu model. The model is built on the GLM-4.1V-Thinking base and has 9B parameters. In tests on the fictional world MagicWorld, the average accuracy of image generation was 54%, while text-only question answering was only 14.57%; on numerical tasks, the two were 72% and 2%, respectively. After interleaved image-text alignment training, text accuracy rose to 26.86%; further adopting a visual chain-of-thought approach in which the model first draws an image and then answers, accuracy could be raised to 41.14%. The research is currently public in the form of a paper, and no subsequent progress has been disclosed.
AI-generated from public reports
Latest update · Discussion
English text is being prepared. joined this discussionDevelopments
1 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 27
English text is being prepared.Related discussion
English text is being prepared. joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Oct 26
36Kr · AI
Tang Jie-authored paper discloses Zhipu's internal multimodal model GLM-Omni
Researchers from Tsinghua University and Zhipu published the paper "Can Visual Generation Expand What Unified Multimodal Models Know?" on arXiv. The corresponding author is Zhipu Chief Scientist Tang Jie. The paper introduces a previously undisclosed Zhipu internal model, GLM-Omni. Built on the GLM-4.1V-Thinking base with 9B parameters, the model achieved an average accuracy of 54% in image generation on the fictional-world MagicWorld test, versus only 14.57% for text-only question answering; on numerical tasks the figures were 72% and 2%, respectively. After interleaved image-text alignment training, text accuracy rose to 26.86%, and a visual chain-of-thought approach in which the model first draws an image and then answers raised it further to 41.14%.