IBM/Dense

IBMgranite-4.0-h-micro 3.2B

chat
3.2B
Parameters
128K
Context length
6
Benchmarks
6
Quantizations
18K
HF downloads
Architecture
Dense
Released
2025-09-16
Layers
40
KV Heads
8
Head Dim
64
Family
granite

Quantization Options

QuantBitsVRAMQuality
Q4_K_M4.892.4 GBgood
Q5_K_S5.572.7 GBgood
Q5_K_M5.72.8 GBgood
Q6_K6.563.1 GBexcellent
Q8_08.53.9 GBlossless
FP16166.9 GBlossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Benchmarks (6)

IFEval51.4
BBH21.7
MMLU-PRO20.2
MATH9.2
GPQA6.6
MUSR1.3

Run this model

Easiest way to get starteddocs →
curl -fsSL https://ollama.com/install.sh | sh
$ollama run granite:3b-q4_k_m

Downloads and runs automatically. Add --verbose for speed stats.

Setup guide

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

Find the best GPU for granite-4.0-h-micro 3.2B

Build Hardware for granite-4.0-h-micro 3.2B

granite-4.0-h-micro 3.2B3.2B Parameter Dense LLM

Model Specifications

Parameters
3.2B
Architecture
Dense Transformer
Context Length
128K tokens
Capabilities
chat
Release Date
2025-09-16
Provider
IBM
Family
granite

VRAM Requirements

QuantizationBPWVRAMQuality
Q4_K_M4.892.4 GB94%
Q5_K_S5.572.7 GB96%
Q5_K_M5.72.8 GB96%
Q6_K6.563.1 GB97%
Q8_08.53.9 GB100%
FP16166.9 GB100%

Benchmark Scores

MMLU-PRO20.2
MATH9.2
IFEval51.4
BBH21.7
GPQA6.6
MUSR1.3

How to Run granite-4.0-h-micro 3.2B

Run granite-4.0-h-micro 3.2B locally with Ollama (needs 2.4 GB VRAM at Q4_K_M):

ollama run granite:3b

Compatible GPUs (30)

GPUs that can run granite-4.0-h-micro 3.2B at Q4_K_M quantization: