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

Russian · 21 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 whisper-large-v3 4.96 2.0× 1.0× 1.55 GB Apache-2.0
2 gigaam-v3-e2e-rnntoverall pickaccurate pickfast pick 5.35 15.0× 7.9× 0.25 GB MIT
3 Voxtral-Mini-4B-Realtime-2602 5.41 0.8× 0.6× 4.41 GB Apache-2.0
4 gigaam-v3-e2e-ctc 5.53 35.4× 11.5× 0.25 GB MIT
5 whisper-large-v2 5.61 2.0× 1.0× 1.55 GB Apache-2.0
6 parakeet-ultra 5.63 23.0× 13.3× 0.69 GB CC-BY-4.0
7 whisper-large-v3-turbo 5.93 3.1× 1.4× 0.83 GB Apache-2.0
8 Qwen3-ASR-1.7B 6.25 4.1× 2.5× 2.04 GB Apache-2.0
9 whisper-large 6.49 2.0× 1.0× 1.55 GB Apache-2.0
10 parakeet-tdt-0.6b-v3 6.54 25.1× 13.7× 0.69 GB CC-BY-4.0
11 orukeet 6.70 25.1× 13.6× 0.69 GB CC-BY-SA-4.0
12 whisper-medium 7.30 4.1× 1.9× 0.77 GB Apache-2.0
13 parakeet-primeline 7.81 25.0× 13.6× 0.69 GB CC-BY-4.0
14 canary-1b-v2! 7.83 14.1× 8.3× 1.07 GB CC-BY-4.0
15 gigaam-v3-rnnt 8.07 26.7× 10.3× 0.25 GB MIT
16 gigaam-v3-ctc 8.42 39.8× 11.8× 0.25 GB MIT
17 Qwen3-ASR-0.6B 10.30 9.7× 5.3× 0.79 GB Apache-2.0
18 whisper-small 11.90 11.7× 5.9× 0.25 GB Apache-2.0
19 nemotron-3.5-asr-streaming-0.6b 12.61 17.1× 10.6× 0.70 GB OpenMDW-1.1
20 whisper-base 22.92 29.0× 19.3× 0.08 GB Apache-2.0
21 whisper-tiny 35.64 43.1× 39.8× 0.04 GB Apache-2.0