List your company
ResearchFormerly WizWand

Automatic Speech Recognition on AISHELL-1 (CER)

0.54CERFireRed-ASR
-0.56846.913314.39521.8767Jan 15, 2026Feb 6, 2026Feb 28, 2026Mar 22, 2026Apr 13, 2026May 5, 2026May 27, 2026
Updated 5mo ago

Evaluation results

MethodLinks
FireRed-ASRParams=1.1B2026.010.54
FireRedASR-AED2026.030.55
Kimi-AudioBase Model=Kimi-Audio2026.020.61
Step-Audio2Params=null2026.010.63
Seed-ASRParams=null2026.010.68
Seed-ASR2026.030.68
Kimi-AudioParams=null2026.010.71
Fun-ASRParams=7.7B2026.011.22
Uni-ASRStreaming mode=false, Beam search decoding width=32026.031.44
Polyglot-Lion-1.7BParams=1.7B2026.031.45
Qwen3-ASR-1.7BModel Scale=1.7B2026.051.5
Qwen3-ASR-1.7BParams=1.7B2026.031.52
Qwen2-AudioBase Model=Qwen2-Audio2026.021.52
Seed-ASR*Params=null2026.011.63
Fun-ASR-nanoParams=0.8B2026.011.8
Fun-ASR-nano2026.031.8
GLM-ASR-nanoParams=1.5B2026.011.81
GLM-ASR-nano2026.031.81
GPA-3BParams=3.0B2026.011.93
Polyglot-Lion-0.6BParams=0.6B2026.031.93
Ark-Base+TD+OPDModel Scale=0.6B, TD=Teacher-data adaptation, OPD=On-policy distillation2026.051.95
SampBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.022.05
Qwen3-ASR-0.6BModel Scale=0.6B2026.052.07
Qwen3-ASR-0.6BParams=0.6B2026.032.08
Uni-ASRStreaming mode=true, Chunk length=1000 ms, Beam search decoding width=32026.032.15
Uni-ASRChunk length=10002026.032.15
GLM-ASR-nano*Params=1.5B2026.012.17
FastAdaSPBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.022.22
GLM-ASR-NanoBase Model=GLM-ASR-Nano2026.022.47
SampBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.022.69
CDPrunerBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.022.7
Ark-Base+OPDModel Scale=0.6B, OPD=On-policy distillation2026.053
MERaLiON-2-10B-ASRParams=10B2026.033.09
FastAdaSPBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.023.28
Ark-BaseModel Scale=0.6B, Checkpoints=100k-hour supervised2026.053.48
SampBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.023.53
A-ToMeBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.023.58
A-ToMeBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.024.18
VisPrunerBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.024.36
GPA-0.3BParams=0.3B2026.014.5
Whisper-LParams=1.55B2026.014.72
CDPrunerBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.024.88
CDPrunerBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.025.07
MoCha-ASR2026.035.1
A-ToMeBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.025.4
FastAdaSPBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.026.42
VisionZipBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.026.84
VisPrunerBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.026.91
VisionZipBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.026.99
Qwen2.5-Omni-7BParams=7B2026.037.33
Whisper-large-v3-turboParams=0.8B2026.039.64
SeaLLMs-Audio-7BParams=7B2026.039.65
VisPrunerBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.0211.68
VisionZipBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.0213.9
Qwen2.5-Omni-3BParams=3B2026.0328.25