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

Portuguese · 25 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 Voxtral-Small-24B-2507 3.74 — — 24.04 GB Apache-2.0
2 Voxtral-Mini-3B-2507 3.84 1.1× 0.6× 4.66 GB Apache-2.0
3 Voxtral-Mini-4B-Realtime-2602 3.87 0.8× 0.6× 4.41 GB Apache-2.0
4 whisper-large-v3 3.88 2.0× 1.0× 1.55 GB Apache-2.0
5 parakeet-ultraoverall pickaccurate pickfast pick 4.16 23.0× 13.3× 0.69 GB CC-BY-4.0
6 whisper-large-v3-turbo 4.17 3.1× 1.4× 0.83 GB Apache-2.0
7 Qwen3-ASR-1.7B 4.37 4.1× 2.5× 2.04 GB Apache-2.0
8 whisper-large-v2 4.40 2.0× 1.0× 1.55 GB Apache-2.0
9 whisper-large 4.49 2.0× 1.0× 1.55 GB Apache-2.0
10 canary-1b-v2! 4.50 14.1× 8.3× 1.07 GB CC-BY-4.0
11 orukeet 4.56 25.1× 13.6× 0.69 GB CC-BY-SA-4.0
12 parakeet-tdt-0.6b-v3 4.96 25.1× 13.7× 0.69 GB CC-BY-4.0
13 whisper-medium 5.07 4.1× 1.9× 0.77 GB Apache-2.0
14 parakeet-primeline 5.17 25.0× 13.6× 0.69 GB CC-BY-4.0
15 cohere-transcribe-03-2026! 5.18 8.1× 4.3× 2.25 GB Apache-2.0
16 granite-speech-4.1-2b-nar! 5.57 4.0× 2.1× 2.33 GB Apache-2.0
17 Qwen3-ASR-0.6B 6.57 9.7× 5.3× 0.79 GB Apache-2.0
18 whisper-small 7.65 11.7× 5.9× 0.25 GB Apache-2.0
19 nemotron-3.5-asr-streaming-0.6b 8.52 17.1× 10.6× 0.70 GB OpenMDW-1.1
20 granite-4.0-1b-speech! 9.60 4.6× 2.8× 2.38 GB Apache-2.0
21 granite-speech-4.1-2b! 9.80 4.6× 2.8× 2.38 GB Apache-2.0
22 granite-speech-4.1-2b-plus! 10.61 4.6× 2.6× 2.18 GB Apache-2.0
23 whisper-base 13.91 29.0× 19.3× 0.08 GB Apache-2.0
24 whisper-tiny 24.07 43.1× 39.8× 0.04 GB Apache-2.0
25 Fun-ASR-MLT-Nano-2512! 28.24 8.8× 5.5× 0.83 GB FunASR Model Open Source License Agreement v1.1