Kimi-Audio-7B-Instruct: benchmark results
Provider: Moonshot. Access: API.
Aggregate Rating 1562 ± 23, rank #1515 of 10916 rated models, from 55 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Audio Event Timing QA - Latest Offset | 60.32 | Accuracy (%; which of four sound events ends last, audio onl | 100 |
| Audio Event Timing QA - Longest Duration | 59.37 | Accuracy (%; which of four sound events lasts longest, audio | 100 |
| SonicBench - Reverberation | 77 | Accuracy (%; reverberation (SRIR room impulse responses); 10 | 100 |
| SonicBench - Duration | 70 | Accuracy (%; sound-event duration (TAU); 100 recognition and | 96.8 |
| SonicBench - Pitch | 83 | Accuracy (%; pitch height (NSynth notes, one-semitone contra | 96.8 |
| SonicBench | 67 | Accuracy (%; mean over 12 physical attributes of 100 recogni | 95.2 |
| SonicBench - Brightness | 81 | Accuracy (%; spectral brightness (NSynth); 100 recognition a | 93.5 |
| SonicBench - Counting | 74 | Accuracy (%; number of sound events (PicoAudio); 100 recogni | 93.5 |
| SonicBench - Loudness | 68 | Accuracy (%; loudness (integrated LUFS contrasts of at least | 93.5 |
| SonicBench - Texture | 97 | Accuracy (%; sound texture (ESC-50); 100 recognition and 100 | 92.2 |
| AOR-Bench - MB-Score | 78.9 | MB-Score (%; harmonic mean of the true-refusal rate on 500 h | 90.9 |
| StanceBench - Social Engagement | 0.73 | AUROC (0-1 scale) of the judge's conversation-level signed s | 87.5 |
Interactive version: theaggregate.ai/model?slug=kimi-audio-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.