FitMyLLM
IBM/Dense

IBMGranite 4.0 3B Vision

granite-vision-3.2-2b is a compact and efficient vision-language model, specifically designed for visual document understanding, enabling automated content extraction from tables, charts, infographics, plots, diagrams, and more.

chatvisionmultilingual
3B
Parameters
128K
Context length
7
Benchmarks
6
Quantizations
0
Architecture
Dense
Released
2025-12-01
Layers
40
KV Heads
8
Head Dim
64
Family
granite

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
Q4_K_M4.89
3.3 GB
2.3 + 0.9 KV
good
Q5_K_S5.57
3.5 GB
2.6 + 0.9 KV
good
Q5_K_M5.7
3.6 GB
2.6 + 0.9 KV
good
Q6_K6.56
3.9 GB
2.9 + 0.9 KV
excellent
Q8_08.5
4.6 GB
3.7 + 0.9 KV
lossless
FP1616
7.4 GB
6.5 + 0.9 KV
lossless

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

Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise →
Ready to run this?Rent by the hour

Rent a GPU and run Granite 4.0 3B Vision now

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (7)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run granite4:3b

Ollama's default build of this model. For Q4_K_M specifically, pick a tag from ollama.com/library/granite4/tags.

Setup guide
>_

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more
Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA Tesla C2050
3 GB VRAM • 144 GB/s
NVIDIA
NVIDIA Tesla M2050
3 GB VRAM • 148 GB/s
NVIDIA
NVIDIA Tesla S2050
3 GB VRAM • 148 GB/s
NVIDIA

Find the best GPU for Granite 4.0 3B Vision

Build Hardware for Granite 4.0 3B Vision
Spec sheet

Granite 4.0 3B Vision — 3B Dense.

Specifications
Parameters
3B
Architecture
Dense Transformer
Context Length
128K tokens
Capabilities
chat, vision, multilingual
Release Date
2025-12-01
Provider
IBM
Family
granite
VRAM requirements
QuantBPWVRAMQuality
Q4_K_M4.892.3 GB94%
Q5_K_S5.572.6 GB96%
Q5_K_M5.72.6 GB96%
Q6_K6.562.9 GB97%
Q8_08.53.7 GB100%
FP16166.5 GB100%
Benchmark scores
MMLU-PRO20.2
MATH15.3
IFEval62.9
BBH21.8
MMMU37.0
GPQA5.3
MUSR4.9
Run command

Run Granite 4.0 3B Vision locally with Ollama — needs 2.3 GB VRAM at Q4_K_M:

$ollama run granite4:3b
Compatible GPUs
30 @ Q4_K_M