Automatic Speech Recognition on LibriSpeech (dev, clean/other)
1.21WER (clean)Step-Audio-2-mini
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Step-Audio-2-mini2026.05 | 1.21 | 2.5 | |
| Kimi-Audio-7B2026.05 | 1.38 | 2.56 | |
| GPT-4o-trans.2026.05 | 1.52 | 3.29 | |
| Qwen3-ASR-1.7B2026.05 | 1.62 | 3.07 | |
| MEGA-ASR2026.05 | 1.62 | 3.21 | |
| MEGA-ASRrouter=true2026.05 | 1.64 | 3.07 | |
| Gemini-3-Flash2026.05 | 1.7 | 3.56 | |
| Whisper Large-v32026.05 | 1.74 | 3.68 | |
| Voxtral-Mini-3B-25072026.05 | 1.89 | 3.88 | |
| Parakeet-TDT-0.6B-v32026.05 | 1.91 | 3.54 | |
| Qwen2.5-Omni-7B2026.05 | 2.05 | 4.19 | |
| Canary-1B-v22026.05 | 2.07 | 4.03 | |
| Doubao-LLM ASR2026.05 | 2.95 | 4.06 |