List your company
ResearchFormerly WizWand

Automatic Speech Recognition on Librispeech

0.0141WERKimi-Audio
-0.0246960.2371770.499050.760923Sep 26, 2024Dec 24, 2024Mar 24, 2025Jun 21, 2025Sep 19, 2025Dec 17, 2025Mar 17, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Kimi-AudioModel Category=Medium Sized Large Audio Language Models, Parameter Count=5B-20B parameters2025.090.0141
Voxtral-Small-24BModel Category=Large Sized Large Audio Language Models, Parameter Count=>20B parameters2025.090.0162
Qwen3-Omni-30B-A3B-ThinkingModel Category=Large Sized Large Audio Language Models, Parameter Count=>20B parameters2025.090.0164
Qwen2.5-Omni-7BModel Category=Medium Sized Large Audio Language Models, Parameter Count=5B-20B parameters2025.090.0174
Phi-4-Multi-modalModel Category=Medium Sized Large Audio Language Models, Parameter Count=5B-20B parameters2025.090.0197
Polyglot-Lion-1.7BParams=1.7B2026.030.021
Voxtral-Mini-3BModel Category=Small-sized Audio Language Models, Parameter Count=<5B parameters2025.090.021
Gemini2.5-FlashModel Category=Proprietary Audio Language Models2025.090.0217
Qwen3-ASR-1.7BParams=1.7B2026.030.0231
MERaLiON-2-10B-ASRParams=10B2026.030.0254
Polyglot-Lion-0.6BParams=0.6B2026.030.0267
Qwen3-ASR-0.6BParams=0.6B2026.030.0274
GLM-4-VoiceModel Type=Speech LLM2026.030.0282
EMOVAModel Size=72B2024.090.029
WhisperModel Size=Large2024.090.03
Whisper-large-v3-turboParams=0.8B2026.030.0304
VITAModel Size=8x7B2024.090.034
EMOVAModel Size=7B2024.090.041
GPT-4o-transcribe + GPT-4.1-miniModel Category=Cascaded Systems2025.090.0471
Mini-Omni22024.090.048
Whisper (No SAE)Steering Method=None, SAE=None2026.020.051
Whisper + AudioSAESteering Method=None, Configuration=Injected SAE on last layer2026.020.052
S-Vector SteeringAlpha (α)=3, Calculation Dataset=Musan2026.020.053
EMOVAModel Size=3B2024.090.054
AudioSAE SteeringAlpha (α)=1, Features=Top-100 from FSD50k2026.020.055
GPT-4o-mini-audioModel Category=Proprietary Audio Language Models2025.090.0625
Omni-DiffusionModel Type=Any-to-Any2026.030.0705
Qwen2.5-Omni-3BModel Category=Small-sized Audio Language Models, Parameter Count=<5B parameters2025.090.0809
VITAModel Size=1.52024.090.081
AnyGPTModel Type=Any-to-Any2026.030.085
Whisper-Large-v3 + GPT-oSS-20BModel Category=Cascaded Systems2025.090.0982
Qwen2.5-Omni-7BParams=7B2026.030.138
Qwen2.5-Omni-3BParams=3B2026.030.2921
SeaLLMs-Audio-7BParams=7B2026.030.9474
AudioSAE SteeringAlpha (α)=3, Features=Top-100 from FSD50k2026.020.984