Project AURIONAI Use-Cases Research & Innovation
IBM
LLMfamily

Granite 4.0

IBM Granite 4.0 family of hybrid Mamba and transformer models (H-Small, H-Tiny, H-Micro) plus a conventional Micro model, released under Apache 2.0.

IBMIBMUSA
Charts of memory needed by Granite 4.0 against other models at long context and many sessions, from the IBM announcement.
Levelfamily
ClassLLM
Parametersnot in the notes
Contextnot in the notes
Released2 Oct 2025
Opennessopen weights permissive
Runs onnot in the notes
Statusnot in the notes

Benchmark Results

vendor-reported
BFCLagents and tool use
Granite-4.0-H-MicroBFCL v3vendor-reported
57.56
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
GPQAknowledge and reasoning
Granite-4.0-H-Micro0-shot, CoTvendor-reported
32.15
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
GSM8Kmath
Granite-4.0-H-Micro8-shotvendor-reported
81.35
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
HumanEvalcoding
Granite-4.0-H-Micropass@1vendor-reported
81
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
IFEvalknowledge and reasoning
Granite-4.0-H-MicroAverage of instruct and prompt strictvendor-reported
84.32
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
MMLUknowledge and reasoning
Granite-4.0-H-Micro5-shotvendor-reported
67.43
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
MMLU-Proknowledge and reasoning
Granite-4.0-H-Micro5-shot, CoTvendor-reported
43.48
050100

Score in %. Higher is better.

Granite-4.0-H-Micro model card2 Oct 2025 huggingface.co
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

Lineage

IBMGranite-4.0-H-Micro belongs to this familyedge LLM2 Oct 20253B
IBMGranite-4.0-H-Micro-Base belongs to this familyLLM
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. Confidence and evidence are the notes' own.

Identity

Developer
IBM
Family
not in the notes
Released
2 Oct 2025 sourcesannouncement page
Weights published
–
Record checked
11 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
not in the notes
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Context length
not in the notes
Input
not in the notes
Output
not in the notes
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights permissive
Licence
Apache License 2.0 sources
Access routes
not in the notes
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

R&D Topics

No R&D topic links this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. ibm.com/new/announcements/ibm-granite-4-0-hyper-effic... https://www.ibm.com/new/announcements/ibm-granite-4-0-hyper-efficient-high-performance-hybrid-models

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Granite 4.0", https://aurion.semai.id/models/ibm-granite-4-0/, records of October 2026.