Automatic Speech Recognition on CommonVoice 13 languages (test)
9.18WERQwen3-ASR
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Qwen3-ASRModel Size=1.7B2026.01 | 9.18 | — | — | — | — | — | — | |
| WhisperVariant=large-v32026.01 | 10.77 | — | — | — | — | — | — | |
| Qwen3-ASRModel Size=0.6B2026.01 | 12.75 | — | — | — | — | — | — | |
| Fun-ASRVariant=MLT-Nano2026.01 | 17.25 | — | — | — | — | — | — | |
| GLM-ASRVariant=Nano-25122026.01 | 19.4 | — | — | — | — | — | — | |
| BottleneckPar (M)=1.22026.06 | — | 54.5 | 50.5 | 22.2 | 60.4 | 65.8 | 50.7 | |
| ConformerPar (M)=1.42026.06 | — | 59.7 | 55.6 | 25.1 | 67.3 | 71 | 55.7 | |
| FFTPar (M)=2412026.06 | — | 46.8 | 46.2 | 18.2 | 53 | 61.7 | 45.18 | |
| LoRAPar (M)=1.22026.06 | — | 62.4 | 57 | 27 | 67.8 | 72.3 | 57.3 | |
| MambAdapterPar (M)=1.12026.06 | — | 53.2 | 50.2 | 22.3 | 58.8 | 64.9 | 49.9 |