MiMo-Audio: benchmark results
Provider: Xiaomi.
Aggregate Rating 1607 ± 32, rank #946 of 10916 rated models, from 80 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| VocalGrad - Sine-Wave Ablation - Noise | 64.2 | Accuracy (%) | 100 |
| VocalGrad - Vibration | 50.3 | Accuracy (%) | 100 |
| VocalGrad - Volume | 52.9 | Accuracy (%) | 100 |
| FMSU-Bench - Acoustic Environment (Chinese) | 75.8 | Accuracy (%) of identifying the acoustic environment inferre | 88.9 |
| MSU-Bench - Accent Identification | 59 | Accuracy (%) on the accent identification task (Tier 1), fou | 87.5 |
| MMAU | 72.59 | Avg Test (imported) | 80 |
| FMSU-Bench - Accent (English) | 64.8 | Accuracy (%) of identifying the accent of the speaker (ACC), | 77.8 |
| FMSU-Bench - Acoustic Environment (English) | 79.2 | Accuracy (%) of identifying the acoustic environment inferre | 77.8 |
| FMSU-Bench - Emotion (English) | 64.3 | Accuracy (%) of identifying the emotion conveyed by the voca | 77.8 |
| MSU-Bench - Speaker Counting | 58 | Accuracy (%) on the speaker counting task (Tier 1), four-opt | 75 |
| ParaIntent - Intent Fulfillment | 47.95 | Intent fulfillment score (%; GLM-5 judge of whether the spok | 75 |
| ParaIntent - Response Quality | 52.91 | Response quality score (%; held-out DeepSeek-V4-Pro judge of | 75 |
Interactive version: theaggregate.ai/model?slug=mimo-audio · How It Works · Data refreshed daily, snapshot 2026-11-08.