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

Turkish · 7 models · left is slower, down is better · log speed axis

AMD Ryzen 4750U GPU (Vulkan)
pareto frontier (no model is both faster and more accurate) other models
1 whisper-large-v3most accurate 6.515.90–7.13 2.1× 0.6× 1.55 GB Apache-2.0
2 whisper-large-v3-turbooverall pick 6.976.33–7.64 3.4× 0.8× 0.83 GB Apache-2.0
3 whisper-medium 9.358.65–10.19 4.3× 1.1× 0.77 GB Apache-2.0
4 Qwen3-ASR-1.7B 9.468.72–10.22 3.8× 2.0× 2.04 GB Apache-2.0
5 nemotron-3.5-asr-streaming-0.6bfast pick 15.4014.46–16.35 14.5× 7.5× 0.70 GB OpenMDW-1.1
6 whisper-small 15.9514.93–17.00 12.1× 3.4× 0.25 GB Apache-2.0
7 Qwen3-ASR-0.6B 16.7415.89–17.74 8.0× 4.3× 0.79 GB Apache-2.0