List your company
ResearchFormerly WizWand

Automatic Speech Recognition on AISHELL D 2021 (test)

1.66CERQwen3-Omni-Inst
1.43122.97564.526.0644Apr 9, 2026Apr 10, 2026Apr 12, 2026Apr 14, 2026Apr 16, 2026Apr 18, 2026Apr 20, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Qwen3-Omni-InstModel Size=30B-A3B (↑)2026.041.66
Qwen3-Omni InstructModel Size=30B-A3B, Inference Mode=Offline2026.041.66
NIM4-ASRModel Size=2.3B, Inference Mode=Offline2026.041.81
Qwen3-ASR-1.7BModel Size=2.0B (↓)2026.041.82
Qwen3-ASR 1.7BModel Size=2.0B, Inference Mode=Offline2026.041.82
NIM4-ASRModel Size=2.3B, Inference Mode=Stream2026.042.22
Fun-ASR-nanoModel Size=0.8B (↓)2026.042.33
Fun-ASR NanoModel Size=0.8B, Inference Mode=Offline2026.042.33
Step-Audio2-miniModel Size=8B+ (↑)2026.042.34
Step-Audio2 MiniModel Size=8B+, Inference Mode=Offline2026.042.34
OursModel Size=2.3B2026.042.53
GLM-ASR-nanoModel Size=1.5B (↓)2026.043.4
GLM-ASR NanoModel Size=1.5B, Inference Mode=Offline2026.043.4
FireRedASR2S LLMModel Size=8B+, Inference Mode=Offline2026.044.68
FireRedASR-LLMModel Size=8B+ (↑)2026.047.38