best model by language

Benchmarked on the FLEURS test split at Q8_0. Error is CER — lower is better. Speed is × realtime on Handy target hardware.

accuracy vs speed

Thai · 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 Qwen3-ASR-1.7Bmost accurate 6.896.15–7.73 3.8× 2.0× 2.04 GB Apache-2.0
2 Fun-ASR-MLT-Nano-2512!overall pickfast pick 7.997.20–8.97 9.0× 4.5× 0.83 GB FunASR Model Open Source License Agreement v1.1
3 whisper-large-v3 8.788.10–9.55 2.1× 0.6× 1.55 GB Apache-2.0
4 Qwen3-ASR-0.6B 8.818.04–9.71 8.0× 4.3× 0.79 GB Apache-2.0
5 whisper-large-v3-turbo 13.1512.45–14.04 3.4× 0.8× 0.83 GB Apache-2.0
6 whisper-medium 16.0815.23–16.99 4.3× 1.1× 0.77 GB Apache-2.0
7 whisper-small 22.5521.71–23.49 12.1× 3.4× 0.25 GB Apache-2.0