Kimi-Audio: benchmark results
Provider: Moonshot. Released 2025-04-25. Access: API.
Aggregate Rating 1547 ± 21, rank #1739 of 10916 rated models, from 195 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SEA-SpeechBench - Emotion Recognition (English Prompt) | 36.86 | Judge-based accuracy (%; closed nine-class emotion label fro | 100 |
| SEA-SpeechBench - Emotion Recognition (SEA Prompt) | 42.55 | Judge-based accuracy (%; closed nine-class emotion label fro | 100 |
| StyleBench (Speaking Style) - Angry (Turn 2) | 68.75 | Valid sample percentage (%): share of the 160 angry-emotion | 100 |
| StyleBench (Speaking Style) - Disgusted (Turn 2) | 76.25 | Valid sample percentage (%): share of the 160 disgusted-emot | 100 |
| StyleBench (Speaking Style) - Fearful (Turn 2) | 55 | Valid sample percentage (%): share of the 160 fearful-emotio | 100 |
| StyleBench (Speaking Style) - Happy (Turn 2) | 47.5 | Valid sample percentage (%): share of the 160 happy-emotion | 100 |
| StyleBench (Speaking Style) - Sad (Turn 2) | 73.13 | Valid sample percentage (%): share of the 160 sad-emotion te | 100 |
| StyleBench (Speaking Style) - Speed (Turn 2) | 81.88 | Valid sample percentage (%): share of the 160 speed test dia | 100 |
| StyleBench (Speaking Style) - Speed (Turn 3) | 78.75 | Valid sample percentage (%): share of the 160 speed test dia | 100 |
| StyleBench (Speaking Style) - Surprised (Turn 2) | 53.75 | Valid sample percentage (%): share of the 160 surprised-emot | 100 |
| VocalGrad - Distortion | 51.6 | Accuracy (%) | 100 |
| VocalGrad - Echo | 54.5 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=kimi-audio · How It Works · Data refreshed daily, snapshot 2026-11-08.