GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

Benchmark Model Rank Results
visual-question-answering-vqa-on-mmbench-cnGLM-4.6V#5Accuracy: 88.2
vision-language-understanding-on-mmbench-v1-1GLM-4.6V#8Accuracy: 88.8
video-question-answering-on-mvbenchGLM-4.6V#11Avg.: 74.9
video-question-answering-on-video-mme-with-subtitlesGLM-4.6V (w/ subtitles)#12Accuracy: 81.8
zero-shot-video-question-answer-on-video-mme-w-o-subsGLM-4.6V#12Accuracy (%): 74.8
visual-question-answering-vqa-on-hallusionbench-average-accuracyGLM-4.6V#13Accuracy: 62.3
multimodal-reasoning-on-mmmuGLM-4.6V#14Accuracy: 76.0
video-question-answering-on-video-mmmuGLM-4.6V#14Accuracy: 74.7
visual-grounding-on-refcoco-refcoco-refcocog-overallGLM-4.6V#14Accuracy: 88.6
video-question-answering-on-mmvuGLM-4.6V#15Accuracy: 68.4
multimodal-reasoning-on-mathvistaGLM-4.6V#16Accuracy: 85.2
video-question-answering-on-lvbenchGLM-4.6V#17Accuracy: 59.5
vision-language-understanding-on-mmstarGLM-4.6V#18Accuracy: 75.9
visual-question-answering-vqa-on-blinkGLM-4.6V#18Accuracy: 65.5
computer-use-agents-on-androidworldGLM-4.6V#22Success Rate: 57.0
multimodal-reasoning-on-mathvisionGLM-4.6V#22Accuracy: 63.5
optical-character-recognition-ocr-on-ocrbenchGLM-4.6V#24Score: 865
multimodal-reasoning-on-mmmu-proGLM-4.6V#27Accuracy: 66.0