Qwen2.5-Omni: benchmark results
Provider: Alibaba.
Aggregate Rating 1558 ± 40, rank #1582 of 10916 rated models, from 77 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HumDial-EIBench - Consistent Emotion (Chinese) | 88 | Accuracy (%) of identifying the speaker's true emotional sta | 100 |
| HumDial-EIBench - Consistent Emotion (English) | 86 | Accuracy (%) of identifying the speaker's true emotional sta | 100 |
| KMMAU - Gender | 98.9 | Accuracy (%; two-way speaker gender question; native Korean | 100 |
| KVoiceBench - KAdvBench | 95.9 | Refusal rate (%; rule-based detection of refusals to harmful | 100 |
| KVoiceBench - KCommonEval | 58.8 | GPT-5.4 judge score (0-100; open-ended spoken requests trans | 100 |
| HumDial-EIBench - Causal Reasoning | 76.99 | Accuracy (%), mean of the Chinese and English accuracies, on | 85.7 |
| KVoiceBench - KOpenBookQA | 31 | Accuracy (%; multiple-choice science questions transferred f | 85.7 |
| DialBGM - Kendall Tau | 0.17 | Kendall tau-b (from -1 to 1) between the model's ranking of | 83.3 |
| SVHalluc - Fine-Grained Semantic Alignment | 81.08 | Accuracy (%; 444 option-balanced video questions, whether a | 80 |
| SVHalluc - Global Semantic Alignment | 77.7 | Accuracy (%; 444 option-balanced video questions, whether th | 80 |
| MusicBench-Global | 78 | Accuracy (%; 2,741 expert-written multiple-choice questions | 75 |
| MMAU | 71 | Avg Test (imported) | 73.3 |
Interactive version: theaggregate.ai/model?slug=qwen2-5-omni · How It Works · Data refreshed daily, snapshot 2026-11-08.