best model by language
Benchmarked on the FLEURS test split at Q8_0. Error is WER — lower is better. Speed is × realtime on Handy target hardware.
accuracy vs speed
Hindi · 15 models
AMD Ryzen 4750U GPU (Vulkan)
pareto frontier (no model is both faster and more accurate) other models
highlighted rows are recommended picks or on pareto frontier
| | | | | | |
| 1 | Voxtral-Small-24B-2507 | 7.40 | — | — | 24.04 GB | Apache-2.0 |
| 2 | Qwen3-ASR-1.7Baccurate pick | 7.84 | 4.1× | 2.5× | 2.04 GB | Apache-2.0 |
| 3 | nemotron-3.5-asr-streaming-0.6boverall pickfast pick | 8.61 | 17.1× | 10.6× | 0.70 GB | OpenMDW-1.1 |
| 4 | Voxtral-Mini-3B-2507 | 8.93 | 1.1× | 0.6× | 4.66 GB | Apache-2.0 |
| 5 | Qwen3-ASR-0.6B | 12.68 | 9.7× | 5.3× | 0.79 GB | Apache-2.0 |
| 6 | Voxtral-Mini-4B-Realtime-2602 | 14.44 | 0.8× | 0.6× | 4.41 GB | Apache-2.0 |
| 7 | whisper-large-v3 | 17.06 | 2.0× | 1.0× | 1.55 GB | Apache-2.0 |
| 8 | whisper-large-v3-turbo | 18.85 | 3.1× | 1.4× | 0.83 GB | Apache-2.0 |
| 9 | whisper-large-v2 | 23.27 | 2.0× | 1.0× | 1.55 GB | Apache-2.0 |
| 10 | whisper-medium | 26.09 | 4.1× | 1.9× | 0.77 GB | Apache-2.0 |
| 11 | whisper-large | 29.10 | 2.0× | 1.0× | 1.55 GB | Apache-2.0 |
| 12 | whisper-small | 42.05 | 11.7× | 5.9× | 0.25 GB | Apache-2.0 |
| 13 | Fun-ASR-MLT-Nano-2512! | 43.96 | 8.8× | 5.5× | 0.83 GB | FunASR Model Open Source License Agreement v1.1 |
| 14 | whisper-base | 100.01 | 29.0× | 19.3× | 0.08 GB | Apache-2.0 |
| 15 | whisper-tiny | 101.52 | 43.1× | 39.8× | 0.04 GB | Apache-2.0 |