Best speech-to-text models
Transcription models by word error rate, lowest first.
ranked by word error rate · 58 models · updated Oct 15 · 51–58
| # | Model | Creator | Word error rate |
|---|---|---|---|
| 51 | Qwen3 ASR Flash, Alibaba | Alibaba | 0.06 |
| 52 | Rev AI | Rev AI | 0.06 |
| 53 | Raon Speech | Krafton | 0.06 |
| 54 | Parakeet TDT 0.6B V2, NVIDIA | NVIDIA | 0.06 |
| 55 | Gradium Speech-to-Text | Gradium | 0.07 |
| 56 | Gemma 4 12B, Google | 0.09 | |
| 57 | Qwen3.5 Omni Flash | Alibaba | 0.14 |
| 58 | Cloud Speech-To-Text (Chirp), Google | 0.31 |
what word error rate means
The share of words a model gets wrong when transcribing. Lower is better, so this board ranks upward from zero.