Contextual Automatic Speech Recognition on ContextASR-Bench
3.3Average Error RateMing-Flash-Omni
Evaluation Results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ming-Flash-Omni2025.10 | 3.3 | 2.91 | 2.53 | 2.13 | 3.5 | 7.6 | 2.08 | 1.13 | 5.41 | 1.16 | 1.77 | 8.3 | 1.04 | |
| Qwen3-ASR2025.10 | 4.23 | 2.95 | 5 | 3.93 | 2.91 | 8.19 | 4.38 | 0.92 | 5.92 | 1.17 | 1.36 | 10.18 | 3.83 | |
| Qwen3-Omni-30B-A3B-Instruct2025.10 | 4.34 | 1.13 | 1.66 | 0.28 | 4.41 | 21.26 | 1.15 | 0.8 | 5.41 | 0.36 | 1.59 | 13.35 | 0.69 | |
| Gemini2.5-Pro2025.10 | 6.3 | 6.65 | 12.82 | 8.86 | 3.59 | 8.18 | 3.53 | 5.37 | 8.35 | 4.09 | 4.03 | 8.96 | 1.18 | |
| Baichuan-Omni-1.52025.10 | 8.17 | 8.16 | 7.69 | 6.53 | 9.91 | 14.4 | 5.54 | 2.98 | 8.39 | 4.71 | 5 | 16.83 | 7.84 | |
| Kimi-Audio2025.10 | 9.24 | 2.9 | 6.68 | 8.01 | 4.67 | 13.5 | 11.31 | 1.95 | 11.13 | 15.28 | 2.9 | 15.91 | 16.68 |