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

Norwegian Bokmal · 5 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 8.197.34–9.09 2.1× 0.6× 1.55 GB Apache-2.0
2 whisper-large-v3-turbooverall pick 9.108.23–10.07 3.4× 0.8× 0.83 GB Apache-2.0
3 whisper-medium 13.6612.66–14.67 4.3× 1.1× 0.77 GB Apache-2.0
4 nemotron-3.5-asr-streaming-0.6bfast pick 19.2417.92–20.58 14.5× 7.5× 0.70 GB OpenMDW-1.1
5 whisper-small 25.5324.20–26.95 12.1× 3.4× 0.25 GB Apache-2.0

Benchmarked but not listed on the model card: whisper-large-v3, whisper-large-v3-turbo, whisper-medium, whisper-small.