README.md

---
base_model:
- vicharai/ViCoder-html-32B-preview
library_name: transformers
tags:
- gguf
- code
- llamacpp
pipeline_tag: text-generation
---
<h1>VICODER HTML 32B PREVIEW QUANTIZATIONS</h1>

## Overview

[`ViCoder-HTML-32B-preview`](https://huggingface.co/vicharai/ViCoder-html-32B-preview) is a powerful AI model designed to generate full websites, including HTML, Tailwind CSS, and JavaScript.


## Model Quantizations

This model comes in several quantizations, each offering a balance of file size and performance. Choose the one that best suits your memory and quality requirements.
| **Quantization**      | **Size (GB)** | **Expected Quality**                                  | **Notes**                                               |
|-----------------------|---------------|--------------------------------------------------------|---------------------------------------------------------|
| **Q8_0**                  | 34.8          | 🟢 *Very good – nearly full precision*                 | 8-bit quantization, very close to full precision for most tasks. |
| **Q6_K**                  | 26.9          | 🟢 *Good – retains most performance*                   | 6-bit quantization, high quality, efficient for most applications. |
| **Q4_K_M**                | 19.9          | 🟡 *Moderate – usable with minor degradation*          | 4-bit quantization, good tradeoff between quality and size. |
| **Q3_K_M**                | 15.9          | 🟠 *Lower – may lose accuracy, better for small RAM*     | 3-bit quantization, lower quality, best for minimal memory use. |

## Features

- **Full Website Generation**: Generates HTML code with Tailwind CSS and JavaScript for modern, responsive websites.
- **Flexible Quantization**: Choose from various quantization models to fit your hardware and performance requirements.
- **Ease of Use**: The model is easy to integrate using [llama.cpp](https://github.com/ggerganov/llama.cpp) and [Ollama](https://github.com/ollama/ollama)