| speech-recognition-on-common-voice-english | Step-Audio 2 mini | #1 | Test WER: 6.76 |
| speech-recognition-on-common-voice-english | Kimi-Audio | #3 | Test WER: 7.83 |
| speech-recognition-on-common-voice-english | Doubao LLM ASR | – | Test WER: 9.20 |
| speech-recognition-on-common-voice-english | GPT-4o Transcribe | – | Test WER: 9.30 |
| speech-recognition-on-common-voice-english | Qwen Omni | – | Test WER: 8.33 |
| speech-recognition-on-common-voice-english | Step-Audio 2 | – | Test WER: 5.95 |
| speech-recognition-on-librispeech-test-clean | Step-Audio 2 mini | #21 | Word Error Rate (WER): 1.33 |
| speech-recognition-on-librispeech-test-clean | Doubao LLM ASR | – | Word Error Rate (WER): 2.92 |
| speech-recognition-on-librispeech-test-clean | GPT-4o Transcribe | – | Word Error Rate (WER): 1.75 |
| speech-recognition-on-librispeech-test-clean | Qwen Omni | – | Word Error Rate (WER): 2.93 |
| speech-recognition-on-librispeech-test-clean | Step-Audio 2 | – | Word Error Rate (WER): 1.17 |
| speech-recognition-on-librispeech-test-other | Step-Audio 2 mini | #23 | Word Error Rate (WER): 2.86 |
| speech-recognition-on-librispeech-test-other | Doubao LLM ASR | – | Word Error Rate (WER): 5.32 |
| speech-recognition-on-librispeech-test-other | GPT-4o Transcribe | – | Word Error Rate (WER): 4.23 |
| speech-recognition-on-librispeech-test-other | Qwen Omni | – | Word Error Rate (WER): 5.07 |
| speech-recognition-on-librispeech-test-other | Step-Audio 2 | – | Word Error Rate (WER): 2.42 |
| audio-understanding-reasoning-on-mmau | Step-Audio-2 | – | Accuracy (Avg, Test-mini): 77.58 |
| audio-understanding-reasoning-on-mmau | Step-Audio-2-mini (8.3B) | – | Accuracy (Avg, Test-mini): 72.73 |