GR

granite-4.0-h-micro

GraniteGGUF by MungertUpdated 2025-10-21
Download on Hugging Face
Parameters3.2B
Context1024K
Layers40
Architecturegranitehybrid
Smallest quant0.0 GB
Quick answer

granite-4.0-h-micro needs about 5.2 GB of memory at Q4_K_M (a 2.0 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 27 quants, from 0.0 GB to 6.4 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 10.7 GB.

Granite‑4.0‑H‑Micro: 3.2B‑parameter long‑context instruction model

Granite‑4.0‑H‑Micro is a 3.2B‑parameter instruct model with a 1,048,576‑token context window. It is finetuned from Granite‑4.0‑H‑Micro‑Base using open source instruction datasets and synthetic data. The model is released under Apache‑2.0.

Good at

  • Long‑context instruction following
  • Tool‑calling
  • Enterprise‑grade safety

Limits

  • No vision support
  • Requires large VRAM for higher precision quants
  • No mixture of experts

Best for

  • Chatbots needing extended context
  • Tool‑integrated applications
  • Enterprise instruction pipelines

Who it’s for. Developers and enterprises that need a compact, long‑context LLM for instruction‑driven tasks. The model is suitable for users with access to llama.cpp and a GPU with sufficient VRAM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 3.2 GB
Fits Get file →
IQ1_S0.8 GB 4.0 GB
Fits Get file →
IQ1_M0.9 GB 4.1 GB
Fits Get file →
IQ2_XXS1.0 GB 4.2 GB
Fits Get file →
IQ2_XS1.1 GB 4.3 GB
Fits Get file →
IQ2_S1.1 GB 4.3 GB
Fits Get file →
IQ2_M1.2 GB 4.3 GB
Fits Get file →
Q2_K_S1.2 GB 4.4 GB
Fits Get file →
Q2_K_M1.2 GB 4.4 GB
Fits Get file →
IQ3_XXS1.4 GB 4.5 GB
Fits Get file →
IQ3_XS1.4 GB 4.6 GB
Fits Get file →
Q3_K_S1.5 GB 4.7 GB
Fits Get file →
IQ3_M1.5 GB 4.7 GB
Fits Get file →
Q3_K_M1.6 GB 4.7 GB
Fits Get file →
IQ4_XS1.8 GB 4.9 GB
Fits Get file →
IQ4_NL1.8 GB 5.0 GB
Fits Get file →
Q4_K_S1.9 GB 5.1 GB
Fits Get file →
Q4_01.9 GB 5.1 GB
Fits Get file →
Q4_K_M2.0 GB 5.2 GB
Fits Get file →
Q4_12.0 GB 5.2 GB
Fits Get file →
Q5_02.3 GB 5.5 GB
Fits Get file →
Q5_K_M2.3 GB 5.5 GB
Fits Get file →
Q5_12.5 GB 5.6 GB
Fits Get file →
Q6_K_M2.7 GB 5.9 GB
Fits Get file →
Q8_03.4 GB 6.6 GB
Fits Get file →
F164.2 GB 7.4 GB
Tight Get file →
BF166.4 GB 9.6 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (0.81 GB) or IQ1_M (0.91 GB). For 12 GB VRAM, IQ2_S (1.09 GB) or Q2_K_M (1.20 GB) are suitable. For 16 GB VRAM, Q3_K_M (1.55 GB) or Q4_K_M (1.97 GB) provide higher precision. Memory estimates are approximate.

What GPU do you need for granite-4.0-h-micro?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is IQ1_S (0.8 GB file, about 4.0 GB in use).
  • 6 GB VRAM: the best quant that fits is Q6_K_M (2.7 GB file, about 5.9 GB in use).
  • 8 GB VRAM: the best quant that fits is F16 (4.2 GB file, about 7.4 GB in use).
  • 12 GB VRAM: the best quant that fits is BF16 (6.4 GB file, about 9.6 GB in use).

Which granite-4.0-h-micro quant should you choose?

Q4_K_M is the usual starting point for granite-4.0-h-micro: it balances file size and quality well. Here it is a 2.0 GB file.

The Q8_0 file is 3.4 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturegranitehybrid
Parameters3.2B
Layers40
Trained context1,048,576 tokens
KV cache (fp16)320 KB per token

Source

Quantized byMungert
FamilyGranite
Licenseapache-2.0
Downloads (HF)1,702
Last updated2025-10-21

Frequently asked questions about granite-4.0-h-micro

How much VRAM does granite-4.0-h-micro need?

About 5.2 GB at Q4_K_M with an 8K context window. That is the 2.0 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run granite-4.0-h-micro on an 8 GB GPU?

Yes, with F16: the 4.2 GB file needs about 7.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of granite-4.0-h-micro GGUF?

It depends on the quant: file is 0.0 GB, IQ1_S is 0.8 GB, IQ1_M is 0.9 GB, IQ2_XXS is 1.0 GB, IQ2_XS is 1.1 GB, IQ2_S is 1.1 GB and 21 more.

What context length does granite-4.0-h-micro support?

The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 2.7 GB at 8K and 10.7 GB at 32K.

Where can I download granite-4.0-h-micro GGUF?

The files are on Hugging Face in the Mungert/granite-4.0-h-micro-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does granite-4.0-h-micro use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Granite‑4.0‑H‑Micro on 8 GB VRAM?

Yes, use a quant such as IQ1_S or IQ1_M, which are under 1 GB file size and require about 4 GB VRAM at 8K context.

What quant gives the best balance of size and performance?

IQ2_S (1.09 GB) or Q2_K_M (1.20 GB) offer a good trade‑off between file size and precision for most users.

Does the model support vision tasks?

No, Granite‑4.0‑H‑Micro does not include vision capabilities.

Similar models

Browse all →

More from Mungert

All by Mungert →

More Granite models

All Granite →

Other tiny models

All tiny models →