Measured model

Measured on a Mac Studio (M5 Max, 36GB)gemma4:26b

Google

On a Mac Studio (M5 Max, 36GB), gemma4:26b answers a simple question in about 3.2 s, gets 85% of the Japanese knowledge questions right, 98% of the math problems, and uses 18.7GB of memory.

Results

MeasureResultAmong measured modelsNote
Reply timeabout 3.2 s18 of 33Until one simple common-sense question is answered (median of 50)
Knowledge (high school to university)85%4 of 33200 questions, ±6, 2 cut off (counted wrong)
Math (word problems)98%3 of 33100 questions, ±5
Memory used18.7GB25 of 33estimated from the file size
EN→JA translation34 pts4 of 30out of 100 (chrF), about 8.5 s a sentence
JA→EN translation58 pts4 of 30out of 100 (chrF), about 8.3 s a sentence

"Among measured models" is the place among the 33 models measured on this site (lower is better for time and memory).

Measured on other Macs

The numbers and places on this page come from the Mac above alone.

Try it on your Mac

  1. Install the Mac version of Ollama from ollama.com and start it.
  2. Type this in Terminal. The first time, the model downloads; then you can chat with it.ollama run gemma4:26b

It used 18.7GB here. To use it comfortably alongside other apps, a Mac with 48GB of memory or more is a good guide (the model then takes about half the memory).

About the model

Maker
Google
Size
25.2B parameters
Quantization
Q4_K_M
Thinks before answering
Yes
Longest input at once
262,144 tokens
Released (Hugging Face)
Mar 11, 2026
Generation speed
103.4 tok/s
Start-up (loading)
about 5.7 s
Measured on
Nov 2, 2026
Ollama
0.34.3
Hugging Face
google/gemma-4-26B-A4B-it
Series
Gemma

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

Models of similar memory

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

Task recommendations and capabilities · Measurements by model