Qwen3-14B measured on a Mac Studio (M5 Max, 36GB) | Reply time, accuracy and memory

Results

Measure	Result	Among measured models	Note
Reply time	about 8 s	29 of 37	Until one simple common-sense question is answered (median of 50)
Knowledge (high school to university)	82%	11 of 37	200 questions, ±6
Math (word problems)	96%	10 of 37	100 questions, ±6
Memory used	10.6GB	19 of 37
EN→JA translation	34 pts	5 of 33	out of 100 (chrF), about 16.2 s a sentence
JA→EN translation	55 pts	14 of 33	out of 100 (chrF), about 13.1 s a sentence

"Among measured models" is the place among the 37 models measured on this site (lower is better for time and memory).

Try it on your Mac

Install the Mac version of Ollama from ollama.com and start it.
Type this in Terminal. The first time, the model downloads; then you can chat with it.ollama run hf.co/Qwen/Qwen3-14B-GGUF:Q4_K_M

It used 10.6GB here. To use it comfortably alongside other apps, a Mac with 24GB of memory or more is a good guide (the model then takes about half the memory).

About the model

Maker: Alibaba
Size: 14.8B parameters
Quantization: Q4_K_M
Thinks before answering: Yes
Longest input at once: 40,960 tokens
Released (Hugging Face): Apr 27, 2025
Generation speed: 40.9 tok/s
Start-up (loading): about 1.1 s
Measured on: Oct 18, 2026
Ollama: 0.34.3
Hugging Face: Qwen/Qwen3-14B
Series: Qwen

What the parts of a name mean (size, quantization and so on): how to read a model name.

Models of similar memory

Ministral-3-14B-Instruct-2512 (9.9GB of memory, knowledge 71%, reply about 0.1 s)
gpt-oss:20b (12.8GB of memory, knowledge 83%, reply about 1.9 s)
NVIDIA-Nemotron-Nano-12B-v2 (8.1GB of memory, knowledge 80%, reply about 6.8 s)
gemma-4-12B-it (8.1GB of memory, knowledge 85%, reply about 7.5 s)
GLM-4-9B-0414 (6.7GB of memory, knowledge 72%, reply about 0.1 s)

Compare the main models on the text page. The series page lists every model measured.