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

Romanian · 17 models

AMD Ryzen 4750U GPU (Vulkan)
pareto frontier (no model is both faster and more accurate) other models
highlighted rows are recommended picks or on pareto frontier
1 canary-1b-v2!overall pickaccurate pickfast pick 6.87 14.1× 8.3× 1.07 GB CC-BY-4.0
2 whisper-large-v3 9.20 2.0× 1.0× 1.55 GB Apache-2.0
3 parakeet-ultra 10.12 23.0× 13.3× 0.69 GB CC-BY-4.0
4 orukeet 10.86 25.1× 13.6× 0.69 GB CC-BY-SA-4.0
5 whisper-large-v3-turbo 10.90 3.1× 1.4× 0.83 GB Apache-2.0
6 parakeet-tdt-0.6b-v3 12.62 25.1× 13.7× 0.69 GB CC-BY-4.0
7 parakeet-primeline 13.80 25.0× 13.6× 0.69 GB CC-BY-4.0
8 whisper-large-v2 17.10 2.0× 1.0× 1.55 GB Apache-2.0
9 whisper-large 17.52 2.0× 1.0× 1.55 GB Apache-2.0
10 Qwen3-ASR-1.7B 20.46 4.1× 2.5× 2.04 GB Apache-2.0
11 whisper-medium 24.17 4.1× 1.9× 0.77 GB Apache-2.0
12 nemotron-3.5-asr-streaming-0.6b 28.28 17.1× 10.6× 0.70 GB OpenMDW-1.1
13 whisper-small 33.88 11.7× 5.9× 0.25 GB Apache-2.0
14 Qwen3-ASR-0.6B 40.65 9.7× 5.3× 0.79 GB Apache-2.0
15 whisper-base 62.16 29.0× 19.3× 0.08 GB Apache-2.0
16 Fun-ASR-MLT-Nano-2512! 74.39 8.8× 5.5× 0.83 GB FunASR Model Open Source License Agreement v1.1
17 whisper-tiny 82.43 43.1× 39.8× 0.04 GB Apache-2.0