Oryx MLLM: On-Demand Spatial-Temporal Understanding at Arbitrary Resolution

Benchmark Model Rank Results
video-question-answering-on-perception-testOyrx (34B)#1Accuracy (Top-1): 71.4
video-question-answering-on-next-qaOryx-1.5(7B)#8Accuracy: 81.8
visual-question-answering-on-docvqa-testOryx-1.5 32B#26ANLS: 0.927
video-question-answering-on-mvbenchOryx(34B)#53Avg.: 64.7
optical-character-recognition-ocr-on-ocrbenchOryx-1.5 32B#58Score: 746
multimodal-reasoning-on-mmmuOryx-1.5 32B#69Accuracy: 56.1