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

Croatian · 9 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 canary-1b-v2!overall pickmost accuratefast pick 8.407.81–9.01 13.2× 6.7× 1.10 GB CC-BY-4.0
2 whisper-large-v3 10.9410.34–11.55 2.1× 0.6× 1.55 GB Apache-2.0
3 whisper-large-v3-turbo 12.5411.83–13.26 3.4× 0.8× 0.83 GB Apache-2.0
4 parakeet-tdt-0.6b-v3 12.5911.85–13.37 12.5× 7.5× 0.72 GB CC-BY-4.0
5 parakeet-primeline 13.6812.94–14.51 12.5× 7.5× 0.72 GB CC-BY-4.0
6 whisper-medium 21.1920.29–22.07 4.3× 1.1× 0.77 GB Apache-2.0
7 nemotron-3.5-asr-streaming-0.6b 26.2125.13–27.33 14.5× 7.5× 0.70 GB OpenMDW-1.1
8 whisper-small 36.0535.00–37.15 12.1× 3.4× 0.25 GB Apache-2.0
9 Fun-ASR-MLT-Nano-2512! 61.0758.92–63.35 9.0× 4.5× 0.83 GB FunASR Model Open Source License Agreement v1.1