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

English · 57 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 Qwen3-ASR-1.7Baccurate pick 3.23 4.1× 2.5× 2.04 GB Apache-2.0
2 Voxtral-Small-24B-2507 3.55 — — 24.04 GB Apache-2.0
3 canary-qwen-2.5b 3.58 4.0× 2.5× 2.61 GB CC-BY-4.0
4 Voxtral-Mini-3B-2507 3.89 1.1× 0.6× 4.66 GB Apache-2.0
5 parakeet-unified-en-0.6boverall pickfast pick 3.99 25.3× 14.0× 0.68 GB CC-BY-4.0
6 whisper-large-v3 4.03 2.0× 1.0× 1.55 GB Apache-2.0
7 parakeet-tdt-0.6b-v2 4.11 27.4× 14.3× 0.68 GB CC-BY-4.0
8 Breeze-ASR-25 4.14 2.0× 1.0× 1.55 GB Apache-2.0
9 granite-speech-4.1-2b! 4.14 4.6× 2.8× 2.38 GB Apache-2.0
10 whisper-large-v2 4.21 2.0× 1.0× 1.55 GB Apache-2.0
11 Qwen3-ASR-0.6B 4.23 9.7× 5.3× 0.79 GB Apache-2.0
12 parakeet-tdt-1.1b 4.24 16.4× 8.8× 1.18 GB CC-BY-4.0
13 parakeet-ultra 4.26 23.0× 13.3× 0.69 GB CC-BY-4.0
14 granite-speech-5.0-470m-turboctc-nc 4.30 22.8× 17.3× 0.47 GB CC-BY-NC-SA-4.0
15 whisper-large-v3-turbo 4.38 3.1× 1.4× 0.83 GB Apache-2.0
16 orukeet 4.43 25.1× 13.6× 0.69 GB CC-BY-SA-4.0
17 canary-1b! 4.44 9.8× 6.5× 1.08 GB CC-BY-NC-4.0
18 parakeet-rnnt-1.1b 4.45 16.1× 8.6× 1.18 GB CC-BY-4.0
19 granite-speech-4.1-2b-plus! 4.46 4.6× 2.6× 2.18 GB Apache-2.0
20 whisper-large 4.46 2.0× 1.0× 1.55 GB Apache-2.0
21 canary-1b-v2! 4.47 14.1× 8.3× 1.07 GB CC-BY-4.0
22 parakeet-rnnt-0.6b 4.57 27.1× 14.8× 0.68 GB CC-BY-4.0
23 granite-speech-5.0-470m-turboctc 4.61 22.9× 17.2× 0.47 GB Apache-2.0
24 whisper-medium 4.64 4.1× 1.9× 0.77 GB Apache-2.0
25 granite-4.0-1b-speech! 4.66 4.6× 2.8× 2.38 GB Apache-2.0
26 parakeet-tdt_ctc-1.1b 4.68 17.0× 9.3× 1.18 GB CC-BY-4.0
27 canary-1b-flash! 4.75 15.5× 8.8× 0.98 GB CC-BY-4.0
28 parakeet-primeline 4.82 25.0× 13.6× 0.69 GB CC-BY-4.0
29 parakeet-tdt-0.6b-v3 4.83 25.1× 13.7× 0.69 GB CC-BY-4.0
30 cohere-transcribe-arabic-07-2026! 4.88 8.3× 4.3× 2.25 GB Apache-2.0
31 whisper-medium.en 4.88 4.4× 2.1× 0.77 GB Apache-2.0
32 Fun-ASR-MLT-Nano-2512! 4.90 8.8× 5.5× 0.83 GB FunASR Model Open Source License Agreement v1.1
33 cohere-transcribe-03-2026! 5.08 8.1× 4.3× 2.25 GB Apache-2.0
34 MOSS-Transcribe-Diarize! 5.13 3.1× 1.6× 0.92 GB Apache-2.0
35 granite-speech-4.1-2b-nar! 5.33 4.0× 2.1× 2.33 GB Apache-2.0
36 Fun-ASR-Nano-2512! 5.49 10.7× 7.9× 0.83 GB FunASR Model Open Source License Agreement v1.1
37 parakeet-ctc-0.6b 5.53 28.6× 15.2× 0.67 GB CC-BY-4.0
38 parakeet-ctc-1.1b 5.61 16.8× 8.9× 1.17 GB CC-BY-4.0
39 canary-180m-flash! 5.98 33.1× 21.7× 0.20 GB CC-BY-4.0
40 parakeet-tdt_ctc-110m 6.11 76.5× 50.0× 0.13 GB CC-BY-4.0
41 whisper-small.en 6.14 12.5× 6.3× 0.25 GB Apache-2.0
42 Voxtral-Mini-4B-Realtime-2602 6.27 0.8× 0.6× 4.41 GB Apache-2.0
43 nemotron-speech-streaming-en-0.6b 6.43 27.1× 13.6× 0.68 GB NVIDIA Open Model License
44 whisper-small 6.51 11.7× 5.9× 0.25 GB Apache-2.0
45 multitalker-parakeet-streaming-0.6b-v1 6.52 26.3× 13.4× 0.81 GB NVIDIA Open Model License
46 SenseVoiceSmall 7.14 27.7× 19.7× 0.24 GB model-license (FunASR MODEL_LICENSE)
47 whisper-base.en 7.60 32.0× 20.6× 0.08 GB Apache-2.0
48 moonshine-streaming-medium 7.87 14.3× 9.7× 0.28 GB MIT
49 nemotron-3.5-asr-streaming-0.6b 7.90 17.1× 10.6× 0.70 GB OpenMDW-1.1
50 moonshine-streaming-small 8.55 21.8× 13.5× 0.18 GB MIT
51 whisper-base 9.88 29.0× 19.3× 0.08 GB Apache-2.0
52 whisper-tiny.en 10.72 47.5× 41.2× 0.04 GB Apache-2.0
53 moonshine-base 12.25 35.6× 23.5× 0.07 GB MIT
54 whisper-tiny 13.84 43.1× 39.8× 0.04 GB Apache-2.0
55 moonshine-tiny 14.13 60.4× 41.3× 0.03 GB MIT
56 moonshine-streaming-tiny 18.18 59.1× 56.3× 0.05 GB MIT
57 medasr 37.48 71.3× 25.9× 0.12 GB Health AI Developer Foundations