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.
modelVoxtral-Small-24B-2507
AMD 4750U GPU—
AMD 4750U CPU—
Q8_0 size24.04 GB
frontier · AMD 4750U GPU0 of 8
verdict: Best in 5 of 8 languages, and tied with the leader in 3 more (overlapping 95% CIs). Never clearly behind the leader anywhere. No speed data on Handy hardware.
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.
Voxtral-Small-24B-2507 errorbest other model errortie: the 95% CIs overlap
quantization · 8 usable languages at Q5_K_M
Q5_K_M is within noise on every one of 8 usable languages: 66% of the Q8_0 file for no measurable accuracy cost.
Q5_K_M file66% of Q8_015.96 GB against 24.04 GB
median error change−3.3%over 8 usable languages
measurably worse0 of 8above the Q8_0 95% confidence interval
worst usable languageHindi7.40 → 7.67 (+0.27 pts)
measurably worse than Q8_0: above its 95% confidence intervalwithin noisetypical noise of one measurement
per-language numbers (8)
Hindi
7.406.65–8.36
7.67+0.27
+3.6%
German
3.292.93–3.65
3.32+0.03
+0.9%
Spanish
2.862.47–3.25
2.83−0.03
−1.0%
French
3.863.39–4.31
3.76−0.10
−2.6%
English
3.553.07–4.10
3.41−0.14
−3.9%
Portuguese
3.743.25–4.26
3.54−0.20
−5.3%
Dutch
5.124.38–5.96
4.81−0.31
−6.1%
Italian
2.692.19–3.34
2.44−0.25
−9.3%
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.