Granite 4.0 H Small
Creator
IBM Research
Released
2025-09-22
Intelligence
10.8
Artificial Analysis Index
Coding
8.5
Artificial Analysis Index
In $/1M
$0.06
input tokens
Out $/1M
$0.25
output tokens
Blended $/1M
$0.11
3:1 blended
Speed
495
tokens / sec
Benchmark breakdown
Independent evaluation scores, normalized to 0–100.
GPQA Diamond
42
Humanity’s Last Exam
4
MMLU-Pro
62
SciCode
21
LiveCodeBench
25
MATH-500—
AIME 2025
14
τ²-Bench (agentic)
17
Terminal-Bench Hard
2
IFBench
31