handy models

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

model detail

Where this model stands against the best model in each language at Q8_0, its speed on an AMD Ryzen 4750U laptop, and what the card says. Add a second model to compare them head to head.

model moonshine-base-zh
AMD 4750U GPU28.0×
AMD 4750U CPU18.3×
Q8_0 size0.07 GB
frontier · AMD 4750U GPU0 of 1

verdict: Behind the leader's 95% CI in its one benchmarked language, Mandarin Chinese (CER). Faster than realtime on the AMD 4750U GPU (28.0×) and AMD 4750U CPU (18.3×).

license: MIT · card languages: 1 · —

standing by language · 1 benchmarked

Each bar compares this model with the best other model in that language, on a log scale: left of center it is ahead, right of center behind, and the other model is named on the row. The two error rates use the legend colors: this model first, the other model second. Faded bars are ties, where the 95% confidence intervals overlap. Sorted strongest standing first. Hover a row for full details.

moonshine-base-zh error best other model error tie: the 95% CIs overlap

quantization · 1 usable language at F32, F16

F32 is within noise on the one usable language: 319% of the Q8_0 file for no measurable accuracy cost. F16 is within noise on the one usable language: 170% of the Q8_0 file for no measurable accuracy cost.

F32 file319% of Q8_00.23 GB against 0.07 GB
median error change−6.4%over 1 usable language
measurably worse0 of 1above the Q8_0 95% confidence interval
worst usable languageMandarin Chinese (CER)17.79 → 16.65 (−1.14 pts)
F16 file170% of Q8_00.12 GB against 0.07 GB
median error change−6.4%over 1 usable language
measurably worse0 of 1above the Q8_0 95% confidence interval
worst usable languageMandarin Chinese (CER)17.79 → 16.65 (−1.14 pts)
measurably worse than Q8_0: above its 95% confidence interval within noise typical noise of one measurement
per-language numbers (1)
Mandarin Chinese (CER) 17.7915.81–19.77 16.65−1.1416.65−1.14 −6.4%

speed · all 73 models on the AMD 4750U GPU

× realtime, log scale. Left of the realtime marker, transcribing takes longer than the recording. Hover a dot for the model.

compare with

moonshine-tiny-zhsame familyQwen3-ASR-1.7Bahead of it in Mandarin ChineseBreeze-ASR-25closest coverage, 1 language in common