Gemma 4 E4B
Reasoning
Creator
Google DeepMind
Released
2026-04-03
Intelligence
12.2
Artificial Analysis Index
Coding
9.4
Artificial Analysis Index
In $/1M
$0.02
input tokens
Out $/1M
$0.10
output tokens
Blended $/1M
$0.04
3:1 blended
Speed
106
tokens / sec
Benchmark breakdown
Independent evaluation scores, normalized to 0–100.
GPQA Diamond
58
Humanity’s Last Exam
4
MMLU-Pro—
SciCode
24
LiveCodeBench—
MATH-500—
AIME 2025—
τ²-Bench (agentic)
21
Terminal-Bench Hard
8
IFBench
44