Gemma

72.5 Better than usual experience score ▲ 1.1 Confidence: High

111 opinions tracked this week · 7-day rolling window

30-day experience score

Version breakdown

Only comments that explicitly name a version contribute to its score.

Gemma 4 31B	62.5−10.0	n=22	Confidence: Medium
Gemma 4 26B A4B	78.5+6.0	n=11	Confidence: Medium
Gemma 3 27B	mentions only	n=1	Confidence: not enough data

79 generic family mentions are included in the family score, not any version score.

By experience dimension

General text	71.9	n=45
Coding	39.6	n=12
Reasoning	55.5	n=11
Image / vision	not enough data	n=2
Video generation	not enough data	n=0
Roleplay / creative	76.3	n=11
Safety & refusals	not enough data	n=1
Speed & latency	74.1	n=21
Local deploy	63.8	n=10

Top positive drivers

speed quality Speed & latency ×16
problem solved General text ×9
competitor flight General text ×5
strong reasoning Reasoning ×4
speed quality Local deploy ×3

Top negative drivers

competitor flight Coding ×3
performance issue Coding ×3
competitor flight General text ×3
hallucination Reasoning ×2
performance issue Speed & latency ×2

What users are saying

Selected from the latest 7 days, prioritizing recency, evidence quality, confidence, and engagement.

🟢 “Efficient research agent brain, creates synthesis from top ranked sources”
via Reddit · General text · 2026-08-26

🟢 “The big gemma4 runs at about half a second while the big qwen is 1 tok/s. gemma4:12b is a little snappier.”
via Reddit · Speed & latency · 2026-08-26

🟢 “you'll probably get 10+ tokens per second”
via Reddit · Speed & latency · 2026-08-26

🟢 “Gemma4 series works best both from usability and capability perspective”
via Reddit · Speed & latency · 2026-08-25