Qwen2.5-Omni Technical Report

Benchmark Model Rank Results
audio-understanding-reasoning-on-mmauQwen2.5-Omni (8.2B)Accuracy (Avg, Test-mini): 71.50
code-generation-on-humanevalQwen2.5-Omni-7B#34Pass@1: 78.7
zero-shot-video-question-answer-on-egoschema-1Qwen2.5-Omni#3Accuracy: 68.6