Granite-4.0-H-Small needs about 25.4 GB of memory at Q4_K_M (a 19.6 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 27 quants, from 8.7 GB (TQ1_0) to 64.4 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 21.5 GB.
32B parameter Granite-4.0-H-Small with long‑context instruction and tool‑calling
Granite-4.0-H-Small is a 32B parameter long‑context instruct model finetuned from Granite‑4.0‑H‑Small‑Base. It supports up to 1,048,576 context tokens and is designed for enterprise AI assistants. The model is released under Apache 2.0 and supports multiple languages.
Good at
- Instruction following
- Tool‑calling
- Multilingual dialogue
Limits
- Large context requires high VRAM
- No vision support
- Requires Apache 2.0 license compliance
Best for
- Enterprise chat assistants
- RAG and retrieval tasks
- Code and function‑calling workflows
Who it’s for. Developers building AI assistants that need robust instruction following and tool‑calling in a multilingual setting. The model is suitable for businesses that require a large‑context model with enterprise‑grade safety and alignment.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 8.7 GB | 14.6 GB | Too big | Get file → | |
| IQ1_S | 9.6 GB | 15.5 GB | Too big | Get file → | |
| IQ1_M | 10.1 GB | 15.9 GB | Too big | Get file → | |
| IQ2_XXS | 10.9 GB | 16.8 GB | Too big | Get file → | |
| IQ2_M | 11.6 GB | 17.4 GB | Too big | Get file → | |
| Q2_K | 12.0 GB | 17.9 GB | Too big | Get file → | |
| Q2_K_L | 12.0 GB | 17.9 GB | Too big | Get file → | |
| Q2_K_XL | 12.4 GB | 18.3 GB | Too big | Get file → | |
| IQ3_XXS | 13.7 GB | 19.6 GB | Too big | Get file → | |
| Q3_K_S | 14.2 GB | 20.1 GB | Too big | Get file → | |
| Q3_K_XL | 14.6 GB | 20.5 GB | Too big | Get file → | |
| Q3_K_M | 15.5 GB | 21.4 GB | Too big | Get file → | |
| IQ4_XS | 17.4 GB | 23.3 GB | Too big | Get file → | |
| IQ4_NL | 18.4 GB | 24.3 GB | Too big | Get file → | |
| Q4_0 | 18.5 GB | 24.3 GB | Too big | Get file → | |
| Q4_K_S | 18.5 GB | 24.4 GB | Too big | Get file → | |
| Q4_K_XL | 18.8 GB | 24.6 GB | Too big | Get file → | |
| Q4_K_M | 19.6 GB | 25.4 GB | Too big | Get file → | |
| Q4_1 | 20.3 GB | 26.2 GB | Too big | Get file → | |
| Q5_K_S | 22.3 GB | 28.2 GB | Too big | Get file → | |
| Q5_K_M | 22.9 GB | 28.8 GB | Too big | Get file → | |
| Q5_K_XL | 23.1 GB | 29.0 GB | Too big | Get file → | |
| Q6_K | 26.5 GB | 32.4 GB | Too big | Get file → | |
| Q6_K_XL | 27.9 GB | 33.7 GB | Too big | Get file → | |
| Q8_0 | 34.3 GB | 40.1 GB | Too big | Get file → | |
| Q8_K_XL | 38.1 GB | 43.9 GB | Too big | Get file → | |
| BF16 | 64.4 GB (2 parts) | 70.3 GB | Too big | Get file → |
For 8‑GB VRAM, use Q4_0 (18.47 GB file, ~24.3 GB VRAM at 8k context). For 16‑GB VRAM, Q4_K_M (19.57 GB file, ~25.4 GB VRAM) or Q5_K_S (22.35 GB file, ~28.2 GB VRAM). For 32‑GB VRAM, Q6_K (26.53 GB file, ~32.4 GB VRAM) or Q6_K_XL (27.87 GB file, ~33.7 GB VRAM).
What GPU do you need for Granite-4.0-H-Small?
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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: the best quant that fits is IQ1_M (10.1 GB file, about 15.9 GB in use).
- 24 GB VRAM: the best quant that fits is IQ4_XS (17.4 GB file, about 23.3 GB in use).
- 32 GB VRAM: the best quant that fits is Q5_K_XL (23.1 GB file, about 29.0 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_K_XL (38.1 GB file, about 43.9 GB in use).
Which Granite-4.0-H-Small quant should you choose?
Q4_K_M is the usual starting point for Granite-4.0-H-Small: it balances file size and quality well. Here it is a 19.6 GB file.
The Q8_0 file is 34.3 GB, about 1.8 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 TQ1_0 at 8.7 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Granite-4.0-H-Small
How much VRAM does Granite-4.0-H-Small need?
About 25.4 GB at Q4_K_M with an 8K context window. That is the 19.6 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-Small on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 14.6 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Granite-4.0-H-Small GGUF?
It depends on the quant: TQ1_0 is 8.7 GB, IQ1_S is 9.6 GB, IQ1_M is 10.1 GB, IQ2_XXS is 10.9 GB, IQ2_M is 11.6 GB, Q2_K is 12.0 GB and 21 more.
What context length does Granite-4.0-H-Small 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 5.4 GB at 8K and 21.5 GB at 32K.
Where can I download Granite-4.0-H-Small GGUF?
The files are on Hugging Face in the unsloth/granite-4.0-h-small-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-Small 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-Small on 8 GB VRAM?
Use the Q4_0 quantized file; it requires about 24.3 GB VRAM at 8k context, so 8 GB VRAM is insufficient.
Does the model support vision tasks?
No, Granite-4.0-H-Small is an instruction model without vision capabilities.
What license applies to this model?
The model is released under the Apache 2.0 license.