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

Bulgarian · 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 9.228.43–10.04 13.2× 6.7× 1.10 GB CC-BY-4.0
2 parakeet-tdt-0.6b-v3 12.8112.01–13.64 12.5× 7.5× 0.72 GB CC-BY-4.0
3 whisper-large-v3 12.8212.00–13.63 2.1× 0.6× 1.55 GB Apache-2.0
4 whisper-large-v3-turbo 13.5812.78–14.38 3.4× 0.8× 0.83 GB Apache-2.0
5 parakeet-primeline 16.3315.43–17.24 12.5× 7.5× 0.72 GB CC-BY-4.0
6 nemotron-3.5-asr-streaming-0.6b 22.0221.07–23.05 14.5× 7.5× 0.70 GB OpenMDW-1.1
7 whisper-medium 23.0022.00–24.03 4.3× 1.1× 0.77 GB Apache-2.0
8 whisper-small 40.3939.26–41.61 12.1× 3.4× 0.25 GB Apache-2.0
9 Fun-ASR-MLT-Nano-2512! 84.9881.46–89.45 9.0× 4.5× 0.83 GB FunASR Model Open Source License Agreement v1.1