Granite32B params9B active (MoE)128k contextpopular

Granite 4 H-Small 32B-A9B locally

Granite 4 H-Small 32B-A9B is an open-weight LLM from the Granite family with 32B parameters. Main use: chat, RAG and general assistance. Detected minimum hardware: TITAN RTX (24 GB).

Technical facts
Parameters32B
Max context128k
Q4_K_M20.1 GB
Q5_K_M24.6 GB
Q835.8 GB
FP1671.5 GB
FamilyGranite
Last sync2026-05-12

Available quantizations

Q4_K_M
20.1GB

Acceptable. Good compromise when VRAM is limited.

Q5_K_M
24.6GB

Good quality. Sweet spot for size and precision.

Q8
35.8GB

Near-FP16 quality. Comfortable for production.

FP16
71.5GB

Reference precision. Maximum quality, doubled VRAM.

Compatible GPUs

GPUs that can run Granite 4 H-Small 32B-A9B on a single card, ranked by VRAM headroom.

Recommended multi-GPU rigs

For Granite 4 H-Small 32B-A9B at higher quantization or with more context, a multi-GPU rig gives more headroom.

Recommended rig

4× GTX 1060 6GB

Granite 4 H-Small 32B-A9B with Ubuntu, vLLM, Open WebUI and the model already downloaded.

Configure

Similar models

VRAM estimates: parameters x bits/8 plus margin. Real performance varies by engine, context length and batch size.
sync: 2026-05-12