Qwen3-Omni: benchmark results
Provider: Alibaba.
Aggregate Rating 1596 ± 31, rank #1050 of 10916 rated models, from 62 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ParaIntent - Intent Fulfillment | 52.27 | Intent fulfillment score (%; GLM-5 judge of whether the spok | 100 |
| ParaIntent - Response Quality | 58.65 | Response quality score (%; held-out DeepSeek-V4-Pro judge of | 100 |
| Vox-Infinity - Personal Monologues (Audio History) | 70.5 | Accuracy (%; open-ended questions on long personal monologue | 100 |
| Vox-Infinity - Personal Monologues (Text History) | 85.5 | Accuracy (%; open-ended questions on long personal monologue | 100 |
| Vox-Infinity - Ultra Multi-Turn (Text History) | 84.8 | Accuracy (%; open-ended questions on ultra multi-turn spoken | 100 |
| MP-Bench - Discussion Appropriateness | 60.16 | Appropriateness accuracy (%; Discussion scenario, correct on | 90 |
| MP-Bench - Understanding Accuracy | 34.44 | Understanding overall accuracy (%; comprehension question an | 90 |
| IMAVB | 46.4 | Balanced accuracy (%; mean of standard vision, standard audi | 87.5 |
| IMAVB (Fixed Option Order) | 45.9 | Balanced accuracy (%; mean of the four split accuracies, 2,0 | 87.5 |
| MP-Bench - Speaker Name Accuracy | 36.11 | Speaker-name accuracy (%; comprehension questions whose answ | 80 |
| SVHalluc - Cross-Modal Semantic Binding | 74.34 | Accuracy (%; 534 option-balanced video questions, whether an | 80 |
| SVHalluc - Cross-Modal Temporal Binding | 61.75 | Accuracy (%; 339 option-balanced video questions, temporal b | 80 |
Interactive version: theaggregate.ai/model?slug=qwen3-omni · How It Works · Data refreshed daily, snapshot 2026-11-08.