---
pipeline_tag: text-generation
inference: false
license: apache-2.0
library_name: transformers
tags:
- language
- granite-3.2
- TensorBlock
- GGUF
base_model: ibm-granite/granite-3.2-8b-instruct-preview
new_version: ibm-granite/granite-3.2-8b-instruct
---

<div style="width: auto; margin-left: auto; margin-right: auto">
<img src="https://i.imgur.com/jC7kdl8.jpeg" alt="TensorBlock" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</div>
<div style="display: flex; justify-content: space-between; width: 100%;">
    <div style="display: flex; flex-direction: column; align-items: flex-start;">
        <p style="margin-top: 0.5em; margin-bottom: 0em;">
            Feedback and support: TensorBlock's  <a href="https://x.com/tensorblock_aoi">Twitter/X</a>, <a href="https://t.me/TensorBlock">Telegram Group</a> and <a href="https://x.com/tensorblock_aoi">Discord server</a>
        </p>
    </div>
</div>

## ibm-granite/granite-3.2-8b-instruct-preview - GGUF

This repo contains GGUF format model files for [ibm-granite/granite-3.2-8b-instruct-preview](https://huggingface.co/ibm-granite/granite-3.2-8b-instruct-preview).

The files were quantized using machines provided by [TensorBlock](https://tensorblock.co/), and they are compatible with llama.cpp as of [commit b4823](https://github.com/ggml-org/llama.cpp/commit/5bbe6a9fe9a8796a9389c85accec89dbc4d91e39).

<div style="text-align: left; margin: 20px 0;">
    <a href="https://tensorblock.co/waitlist/client" style="display: inline-block; padding: 10px 20px; background-color: #007bff; color: white; text-decoration: none; border-radius: 5px; font-weight: bold;">
        Run them on the TensorBlock client using your local machine ↗
    </a>
</div>

## Prompt template

```
<|start_of_role|>system<|end_of_role|>{system_prompt}<|end_of_text|>
<|start_of_role|>user<|end_of_role|>{prompt}<|end_of_text|>
<|start_of_role|>assistant<|end_of_role|>
```

## Model file specification

| Filename | Quant type | File Size | Description |
| -------- | ---------- | --------- | ----------- |
| [granite-3.2-8b-instruct-preview-Q2_K.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q2_K.gguf) | Q2_K | 3.104 GB | smallest, significant quality loss - not recommended for most purposes |
| [granite-3.2-8b-instruct-preview-Q3_K_S.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q3_K_S.gguf) | Q3_K_S | 3.592 GB | very small, high quality loss |
| [granite-3.2-8b-instruct-preview-Q3_K_M.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q3_K_M.gguf) | Q3_K_M | 3.997 GB | very small, high quality loss |
| [granite-3.2-8b-instruct-preview-Q3_K_L.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q3_K_L.gguf) | Q3_K_L | 4.349 GB | small, substantial quality loss |
| [granite-3.2-8b-instruct-preview-Q4_0.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q4_0.gguf) | Q4_0 | 4.651 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
| [granite-3.2-8b-instruct-preview-Q4_K_S.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q4_K_S.gguf) | Q4_K_S | 4.686 GB | small, greater quality loss |
| [granite-3.2-8b-instruct-preview-Q4_K_M.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q4_K_M.gguf) | Q4_K_M | 4.943 GB | medium, balanced quality - recommended |
| [granite-3.2-8b-instruct-preview-Q5_0.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q5_0.gguf) | Q5_0 | 5.647 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
| [granite-3.2-8b-instruct-preview-Q5_K_S.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q5_K_S.gguf) | Q5_K_S | 5.647 GB | large, low quality loss - recommended |
| [granite-3.2-8b-instruct-preview-Q5_K_M.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q5_K_M.gguf) | Q5_K_M | 5.797 GB | large, very low quality loss - recommended |
| [granite-3.2-8b-instruct-preview-Q6_K.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q6_K.gguf) | Q6_K | 6.705 GB | very large, extremely low quality loss |
| [granite-3.2-8b-instruct-preview-Q8_0.gguf](https://huggingface.co/tensorblock/granite-3.2-8b-instruct-preview-GGUF/blob/main/granite-3.2-8b-instruct-preview-Q8_0.gguf) | Q8_0 | 8.684 GB | very large, extremely low quality loss - not recommended |


## Downloading instruction

### Command line

Firstly, install Huggingface Client

```shell
pip install -U "huggingface_hub[cli]"
```

Then, downoad the individual model file the a local directory

```shell
huggingface-cli download tensorblock/granite-3.2-8b-instruct-preview-GGUF --include "granite-3.2-8b-instruct-preview-Q2_K.gguf" --local-dir MY_LOCAL_DIR
```

If you wanna download multiple model files with a pattern (e.g., `*Q4_K*gguf`), you can try:

```shell
huggingface-cli download tensorblock/granite-3.2-8b-instruct-preview-GGUF --local-dir MY_LOCAL_DIR --local-dir-use-symlinks False --include='*Q4_K*gguf'
```