List your company
ResearchFormerly WizWand

Automatic Speech Recognition on FLEURS en

2.72WERQwen3-Omni-A3B-Instruct
2.3085.0897.8710.651Feb 7, 2026Feb 24, 2026Mar 13, 2026Mar 31, 2026Apr 17, 2026May 4, 2026May 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen3-Omni-A3B-InstructVariant=Omni-A3B-Instruct2026.032.72
Step-Audio-2-miniVersion=2-mini2026.033.05
Qwen3-ASR-1.7B2026.053.23
Qwen3-ASR-1.7B2026.033.35
Seed-ASR2026.033.43
MiniCPM-o 4.52026.043.52
StepAudio 2.5 ASRMTP training=true2026.053.55
Raon-Speech2026.043.59
StepAudio 2.5 ASRMTP training=false2026.053.74
Qwen2.5-Omni2026.044.05
GLM-ASR-NanoBase Model=GLM-ASR-Nano2026.024.11
Whisper-large-v22026.034.2
Uni-ASRStreaming mode=false, Beam search decoding width=32026.034.44
Kimi-Audio2026.034.44
Kimi Audio2026.044.54
Step-Audio Flamingo 32026.044.54
SampBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.024.83
Fun-Audio Omni2026.044.89
Kimi-AudioBase Model=Kimi-Audio2026.024.92
FunASR-Nano2026.054.96
Qwen2-AudioBase Model=Qwen2-Audio2026.025.2
VibeVoice-ASR2026.055.2
LongCat-Next2026.035.24
HyperCLOVA X 8B Omni2026.045.57
A-ToMeBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.025.77
CDPrunerBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.025.93
SampBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.025.94
Fun-ASR-nano2026.035.96
Gemini-3.1-Flash-Lite-previewVersion=3.1, Tier=Flash-Lite-preview2026.035.96
Uni-ASRStreaming mode=true, Chunk length=1000 ms, Beam search decoding width=32026.036.24
SampBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.026.26
MiMo-Audio2026.036.33
FastAdaSPBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.026.7
Doubao-ASR-26032026.056.74
Gemini-2.5-Flash-Lite-previewVersion=2.5, Tier=Flash-Lite-preview2026.036.84
VisionZipBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.026.85
CDPrunerBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.027.17
FastAdaSPBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.027.31
A-ToMeBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.027.57
Fun-Audio Chat2026.047.61
FastAdaSPBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.027.64
VisPrunerBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.027.82
VisionZipBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.028.08
A-ToMeBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.028.13
VisPrunerBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.028.33
VisPrunerBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.028.39
CDPrunerBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.028.66
VisionZipBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.028.91
Interactive 2 mini2026.0413.02