IBM32B (9B active)~20 GB VRAM at Q4_K_M

Granite 4.0 Small-H 32B-A9B — VRAM, speed & local setup

Written by Jakub Rusinowski · Last updated

The largest Granite 4.0 build — 32B total, 9B active, on a hybrid mamba-2 architecture (the '-h' suffix). Also published as `granite4:small-h`. The mamba-2 hybrid keeps memory growth flatter than a pure transformer at long context, which is the point for document-heavy enterprise RAG. Around 19 GB at the default Q4_K_M, so a single 24GB card. Twelve languages including English, German, Spanish, French, Japanese, Portuguese, Arabic, Czech, Italian, Korean, Dutch and Chinese.

Granite 4.0 Small-H 32B-A9B needs about 20 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.

Logic84
Creative74
Coding80

VRAM and speed by quantization

Quoted against NVIDIA RTX 4090 (24 GB). Weights plus overhead only: this model's architecture is unpublished, so no KV cache is added.

QuantVRAMSpeed (est.)Fit
Q2_K
2.63 bpw
11.3 GB~135 tok/sFits
Q3_K_M
3.41 bpw
14.4 GB~117 tok/sFits
Q4_K_M
4.83 bpw
20.1 GB~94 tok/sFits
Q5_K_M
5.67 bpw
23.5 GB~84 tok/sTight
Q6_K
6.56 bpw
27 GB~11 tok/sOffload
Q8_0
8.50 bpw
34.8 GB~9 tok/sOffload
F16
16.00 bpw
64.8 GB—Too big

Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Granite 4.0 Small-H 32B-A9B VRAM calculator →

Get Granite 4.0 Small-H 32B-A9B running

The cheapest catalogued GPU that runs Granite 4.0 Small-H 32B-A9B is the AMD Radeon RX 7900 XTX (24 GB).

Buy this hardware AMD Radeon RX 7900 XTX 24GB — 24 GB VRAM · 355 W board powerDeploy in the cloud now RTX 4090 on RunPod — from $0.34/hr · rate checked 2026-07

or compare on Vast.ai from $0.35/hr (typical low · varies)

As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.

Affiliate disclosure: Some links on this page are affiliate links — if you buy through them, LLM Configurator may earn a commission at no extra cost to you. As an Amazon Associate, LLM Configurator earns from qualifying purchases.
AMD Radeon RX 7900 XTX 24GB
24 GB VRAM · 355 W board power
2026 prices are volatile — check the current listing.

How to run Granite 4.0 Small-H 32B-A9B

Install Ollama, then run:

ollama run granite4:32b-a9b-h
Weights on Hugging Face: ibm-granite/granite-4.0-h-small ↗

Specifications

Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.

Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.

Parameters
32 Billion (9B active)
Context window
128,000
Architecture
Hybrid Mamba-2 + Mixture-of-Experts
Provider
IBM
Licence
Apache 2.0
Practical deployment
Single consumer GPU
Specified at
Q4_K_M
System RAM
32 GB
Record updated
2026-08-15
LicenceApache-2.0Commercial use permitted

Commercial use permitted. No usage restrictions beyond attribution.

Quality and use cases

Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.

Best forenterpriseragtool callingmultilingualon premise

Can I run Granite 4.0 Small-H 32B-A9B on my GPU?

Granite 4.0 Small-H 32B-A9B — frequently asked questions

How much VRAM does Granite 4.0 Small-H 32B-A9B need?

About 20 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.

Does Granite 4.0 Small-H 32B-A9B run on an RTX 4090 (24 GB)?

Yes. Granite 4.0 Small-H 32B-A9B needs about 20 GB at Q4_K_M, inside a 24 GB card, at an estimated 94 tokens/sec.

How do I run Granite 4.0 Small-H 32B-A9B locally?

Install Ollama and run `ollama run granite4:32b-a9b-h`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.