handy models

speech-to-text benchmarks for transcribe.cpp, the engine behind Handy

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

Hungarian · 11 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 13.0612.39–13.80 13.2× 6.7× 1.10 GB CC-BY-4.0
2 whisper-large-v3 13.4012.72–14.14 2.1× 0.6× 1.55 GB Apache-2.0
3 whisper-large-v3-turbo 15.0714.25–15.92 3.4× 0.8× 0.83 GB Apache-2.0
4 parakeet-tdt-0.6b-v3 16.0615.24–16.89 12.5× 7.5× 0.72 GB CC-BY-4.0
5 parakeet-primeline 17.5216.70–18.30 12.5× 7.5× 0.72 GB CC-BY-4.0
6 whisper-medium 26.0725.08–27.04 4.3× 1.1× 0.77 GB Apache-2.0
7 nemotron-3.5-asr-streaming-0.6b 32.1230.89–33.35 14.5× 7.5× 0.70 GB OpenMDW-1.1
8 Qwen3-ASR-1.7B 32.8431.69–33.99 3.8× 2.0× 2.04 GB Apache-2.0
9 whisper-small 42.3941.32–43.49 12.1× 3.4× 0.25 GB Apache-2.0
10 Qwen3-ASR-0.6B 56.2454.94–57.67 8.0× 4.3× 0.79 GB Apache-2.0
11 Fun-ASR-MLT-Nano-2512! 113.21110.70–116.12 9.0× 4.5× 0.83 GB FunASR Model Open Source License Agreement v1.1