Automatic Speech Recognition on Fleurs zh
0.83WERERNIE 5.0
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ERNIE 5.02026.02 | 0.83 | — | |
| ERNIE 5.0-Base2026.02 | 1.58 | — | |
| Qwen3-Omni-Instruct2026.02 | 2.2 | — | |
| Qwen3 Omni2025.10 | 2.2 | — | |
| Qwen3-Omni-A3B-InstructVariant=Omni-A3B-Instruct2026.03 | 2.2 | — | |
| GPT-4o-Audio2026.02 | 2.44 | — | |
| Step-Audio-2-miniVersion=2-mini2026.03 | 2.53 | — | |
| StepAudio 2.5 ASRMTP training=true2026.05 | 2.63 | — | |
| StepAudio 2.5 ASRMTP training=false2026.05 | 2.63 | — | |
| Kimi Audio2026.02 | 2.69 | — | |
| Kimi Audio2025.10 | 2.69 | — | |
| Kimi-Audio2026.03 | 2.69 | — | |
| Qwen3-ASR-1.7B2026.05 | 2.74 | — | |
| Ming-Flash-Omni2025.10 | 2.82 | — | |
| Doubao-ASR-26032026.05 | 2.83 | — | |
| Kimi-Audio-7BReference audio=None2026.04 | 2.87 | — | |
| Qwen3-Omni-30B-A3BReference audio=None2026.04 | 2.88 | — | |
| Qwen2.5-Omni-7BReference audio=None2026.04 | 2.92 | — | |
| UAF-30B-A3BReference audio=None2026.04 | 2.92 | — | |
| FunASR-Nano2026.05 | 3.19 | — | |
| LongCat-Next2026.03 | 3.24 | — | |
| Qwen2-Audio-7BReference audio=None2026.04 | 3.63 | — | |
| Gemini-3.1-Flash-Lite-previewVersion=3.1, Tier=Flash-Lite-preview2026.03 | 3.88 | — | |
| LongCat-Flash-Omni2026.02 | 3.99 | — | |
| Qwen-TTS-Tokenizer-25Hz (Stage 1)Codebook Size=32768, FPS=252026.01 | 4.23 | — | |
| S3 Tokenizer(FSQ)Codebook Size=6561, FPS=252026.01 | 4.43 | — | |
| Qwen-TTS-Tokenizer-25Hz (Stage 2)Codebook Size=32768, FPS=252026.01 | 4.67 | — | |
| Gemini-3-Pro2026.02 | 4.98 | — | |
| S3 Tokenizer(VQ)Codebook Size=4096, FPS=252026.01 | 5.03 | — | |
| Paraformer-zh-streamingReference audio=None2026.04 | 5.98 | — | |
| Qwen2 Audio2025.10 | 7.5 | — | |
| Gemini-2.5-Flash-Lite-previewVersion=2.5, Tier=Flash-Lite-preview2026.03 | 7.77 | — | |
| VibeVoice-ASR2026.05 | 8.77 | — | |
| MiMo-Audio2026.03 | 9.29 | — | |
| A-ToMeBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 8 | |
| A-ToMeBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 6.26 | |
| A-ToMeBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.02 | — | 5.08 | |
| CDPrunerBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 4.88 | |
| CDPrunerBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 6.73 | |
| CDPrunerBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.02 | — | 4.99 | |
| FastAdaSPBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 5.47 | |
| FastAdaSPBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 4.27 | |
| FastAdaSPBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.02 | — | 4.34 | |
| Fun-ASR-nano2026.03 | — | 2.56 | |
| GLM-ASR-NanoBase Model=GLM-ASR-Nano2026.02 | — | 3.44 | |
| Kimi-AudioBase Model=Kimi-Audio2026.02 | — | 2.87 | |
| Qwen2-AudioBase Model=Qwen2-Audio2026.02 | — | 3.63 | |
| Qwen3-ASR-1.7B2026.03 | — | 2.41 | |
| SampBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 4.37 | |
| SampBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 4.37 | |
| SampBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.02 | — | 3.76 | |
| Uni-ASRStreaming mode=false, Beam search decoding width=32026.03 | — | 2.57 | |
| Uni-ASRStreaming mode=true, Chunk length=1000 ms, Beam search decoding width=32026.03 | — | 3.43 | |
| VisionZipBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 8.02 | |
| VisionZipBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 7.63 | |
| VisionZipBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.02 | — | 8.17 | |
| VisPrunerBase Model=Qwen2-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 7 | |
| VisPrunerBase Model=Kimi-Audio, Token Retention Rate=60%, Compression Ratio=40%2026.02 | — | 5.73 | |
| VisPrunerBase Model=GLM-ASR-Nano, Token Retention Rate=70%, Compression Ratio=30%2026.02 | — | 6.95 | |
| Whisper-large-v22026.03 | — | 14.7 |