Granite 4.0 H Small

Granite 4.0 H Small

IBM · Released Sep 22, 2025 Efficient
Intelligence #270 / 599
33.5 Our Score
Speed #3 / 281
428.3 tokens / sec
Input #168 / 599
$0.060 per 1M tokens
Output #205 / 599
$0.250 per 1M tokens
Context
— Not reported

Analysis Summary

Granite 4.0 H Small is IBM's mid-tier small model in the Granite 4.0 family, priced at $0.06/$0.25 per million tokens. Its intelligence index of 5.2 and MMLU-Pro of 0.624 show reasonable general knowledge for its size, and livecodebench at 0.251 gives it basic coding utility.

For businesses, it fits lightweight automation tasks: form processing, short-form content drafting, and simple classification pipelines where cost per call matters. Agentic capability (index 9.8) and long-context reasoning (lcr 0.09) are limited, so it is not suitable for multi-step workflows or document-heavy analysis.

At its price point it competes with other small open-weight models. IBM's enterprise positioning may appeal to organisations already in the IBM ecosystem, but teams without that context should compare it carefully against similarly priced alternatives before adopting.

Assessed June 30, 2026

Editorial notes

Granite 4.0 H Small from IBM is a compact, affordably priced model with moderate instruction following and light coding capability, suited to cost-sensitive automation of simple structured tasks.

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?

Performance Profile

Intelligence1.6Technical1.3Value7Content4.5
Intelligence 1.6/10
Technical 1.3/10
Content 4.5/10
Value 7/10

How Granite 4.0 H Small compares

Granite 4.0 H Small ranks #317 of 387 AI models we track for overall intelligence, #264 of 294 for agentic tasks. At $0.06 per million input tokens it is cheaper than 72% of comparable models.

Performance Indices

Source: Artificial Analysis

5.2 Intelligence Index
9.8 Agentic Index
13.7 Math Index

Benchmark Scores

Intelligence

GPQA Diamond 41.6% Graduate-level scientific reasoning
HLE 3.7% Humanity's Last Exam
MMLU Pro 62.4% Multi-task language understanding
AIME 2025 13.7% Competition mathematics (2025)
SciCode 20.9% Scientific computing

Technical

LiveCodeBench 25.1% Live coding evaluation
TerminalBench Hard 2.3% Agentic terminal tasks
τ²-Bench 17.3% Conversational agent benchmark

Content

IFBench 31.5% Instruction following
LCR 9% Long-context reasoning

Benchmark data from Artificial Analysis and Hugging Face

How does Granite 4.0 H Small stack up?

Compare side-by-side with other efficient models.

Compare Models

Model Information

ProviderIBM
Release Date September 22, 2025
Status Active

Pricing

Token Type Cost per 1M tokens Cost per 1K tokens
Input $0.06 $0.000060
Output $0.25 $0.000250

Frequently asked questions about Granite 4.0 H Small

How much does Granite 4.0 H Small cost?

Granite 4.0 H Small costs $0.06 per million input tokens and $0.25 per million output tokens.

Who created Granite 4.0 H Small?

Granite 4.0 H Small is developed by IBM and was released on September 22, 2025.