Speech Recognition on Aura-STT-BenchLF (DA)
38.48WERGemini 2.5 Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 2.5 Flash2026.03 | 38.48 | |
| Aura-STT*Params=446M, configuration=VAD + Aura-STT (short-form), beam-size=1 (greedy search)2026.03 | 51.28 | |
| Whisper-large-v3Params=1.55B2026.03 | 52.3 | |
| Aura-STT-LF2026.03 | 58.99 | |
| Aura-STT-LFParams=479M2026.03 | 59.5 | |
| Nvidia-Conformer-CTCParams=121M2026.03 | 68.05 | |
| OWSM-CTC-v4Params=1.0B2026.03 | 88.87 | |
| Whisper-v32026.03 | 93 |