Ming-Lite-Omni-1.5: benchmark results

Provider: Other. Access: Open.

Unified ELO 1493 ± 20, rank #766 of 1605 rated models, from 26 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SonicBench - Counting65Accuracy (%; number of sound events (PicoAudio); 100 recogni80.6
SonicBench - Texture94Accuracy (%; sound texture (ESC-50); 100 recognition and 10079.7
SonicBench - Duration53Accuracy (%; sound-event duration (TAU); 100 recognition and69.4
SonicBench57Accuracy (%; mean over 12 physical attributes of 100 recogni67.7
SonicBench - Loudness53Accuracy (%; loudness (integrated LUFS contrasts of at least62.9
SonicBench - Reverberation53Accuracy (%; reverberation (SRIR room impulse responses); 1062.9
SonicBench - Timbre57Accuracy (%; instrument timbre (NSynth); 100 recognition and62.9
SonicBench - Pitch57Accuracy (%; pitch height (NSynth notes, one-semitone contra58.1
MOV-Bench - Hypothetical Reasoning35Accuracy (%; 60 hypothetical-reasoning questions; direct inf57.1
MOV-Bench - Intent Reasoning50Accuracy (%; 80 intent-reasoning questions; direct inference57.1
SonicBench - Brightness52Accuracy (%; spectral brightness (NSynth); 100 recognition a50
MOV-Bench49.52Accuracy (%; 519 audio-visual multi-hop multiple-choice ques42.9

Interactive version: theaggregate.ai/model?slug=ming-lite-omni-1-5 · How It Works · Data refreshed daily, snapshot 2026-09-26.