The LLM benchmark aggregator.

Gemma 4 12B

Google · 12B

Strongest in Secure code (#53 of 268), weakest in Fairness (#246 of 290). Above par in 6 of 16 scopes.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Agents46.7−19.6#186/2981/6

Knowledge work43.4−27.3#132/2081/2

Safety49.1−12.3#203/3271/3

Secure code61.2−5#53/2681/1

Toxicity avoidance54.5−4.5#119/2661/1

Harm refusal48.6−12.9#173/2901/2

Jailbreak resistance40.4−24.6#211/2661/1

Fairness42.3−27.7#246/2901/2

Core abilities47.8−17.8—/200✱0/4

Long context41.7−22.9—/0✱✱

Reasoning52.6−12.8—/195✱0/5

Mathematics53.9−9.9—/152✱0/2

Expert reasoning60.3−6.1—/0✱✱

Coding48.8−17.3—/179✱0/6

Agentic coding47.7−19—/173✱0/5

Code generation51.4−13.9—/20✱0/1

§ 3 · Sources

Where the numbers come from

3 publications, 13 figures. Every one links to the page it was read from.

GDPval-AA ↗1 measureread 2026-11-11

GDPval-AA 435

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-11-11

Enkrypt · Jailbreak risk 20.9%Enkrypt · Harmful content risk 4.4%Enkrypt · CBRN risk 34.5%Enkrypt · Toxicity risk 3.1%Enkrypt · Bias risk 89.7%Enkrypt · Insecure code risk 10.2%

✱✱ IFM — K2 Horizon launch ↗the model’s own publisher6 measuresread 2026-11-03

Terminal-Bench 2.1 27.3%SciCode 38.2%AA-LCR 61.7%SWE-bench Verified 30.6%HMMT Feb 2026 63.1%HLE 15.7%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=gemma-4-12b)](https://publicai.io/model-index/m/gemma-4-12b)