List your company
ResearchFormerly WizWand

Automatic Speech Recognition on AISHELL-2 (ios)

2.29CERStepAudio 2.5 ASR
2.17762.93633.6954.4537Feb 15, 2026Mar 3, 2026Mar 19, 2026Apr 4, 2026Apr 20, 2026May 6, 2026May 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
StepAudio 2.5 ASRMTP training=true2026.052.29
StepAudio 2.5 ASRMTP training=false2026.052.3
Step-Audio-2-miniType=Audio, Size=8B2026.022.33
Qwen3-ASR-1.7B2026.052.5
Qwen2.5-Omni-7BType=Omni, Size=7B2026.022.58
FunASR-Nano2026.052.61
Qwen3-Omni-InstructType=Omni, Size=30B-A3B2026.022.63
Ming-Lite-Omni-1.5Type=Omni, Size=19B-A2.8B2026.022.66
Doubao-ASR-26032026.052.7
Qwen2.5-Omni-3BType=Omni, Size=3B2026.022.77
Kimi-Audio-7B-InstructType=Audio, Size=7B2026.022.8
Qwen2-AudioType=Audio, Size=7B2026.023.08
Eureka-Audio-InstructType=Ours, Size=1.7B2026.023.1
Eureka-Audio-BaseType=Ours, Size=1.7B2026.023.17
MiniCPM-oType=Omni, Size=9B2026.023.42
VibeVoice-ASR2026.055.1