---
base_model: Maxilicious20/Aether-2.5-Coder-3B
library_name: gguf
pipeline_tag: text-generation
tags:
- base_model:adapter:Qwen/Qwen2.5-Coder-3B-Instruct
- gguf
- quantized
- llama-cpp
- coding
- code-generation
- german
- english
- aether
license: apache-2.0
language:
- de
- en
---

# Aether 2.5 Coder - GGUF

This repository contains the quantized GGUF format binaries for **Aether 2.5 Coder**, fine-tuned from **Qwen2.5-Coder-3B-Instruct**. These files are optimized for local inference using tools like **LM Studio**, **Ollama**, **Jan**, **text-generation-webui**, or **llama.cpp**.

> 🔗 **Original LoRA / Adapter Repository:**
> 👉 **[Maxilicious20/Aether-2.5-Coder](https://huggingface.co/Maxilicious20/Aether-2.5-Coder)**

---

## 📦 Available Files

| File Name | Quant Method | File Size | Description |
| :--- | :--- | :--- | :--- |
| `aether_coder_q4_k_m.gguf` | Q4_K_M | ~2.0 GB | **Recommended:** Best balance between performance, low VRAM usage, and speed. |
| `aether_coder_q8_0.gguf` | Q8_0 | ~3.4 GB | **High Quality:** Extremely close to 16-bit precision with minimal loss in accuracy. |
| `aether_coder_f16.gguf` | F16 | ~6.2 GB | **Uncompressed:** Unquantized full-precision GGUF export. |

---

## 🚀 How to Use

### 1. LM Studio / Jan / Local WebUIs
1. Download `aether_coder_q4_k_m.gguf` or `aether_coder_q8_0.gguf`.
2. Move the `.gguf` file into your local models folder.
3. Select the model and start chatting or coding!

### 2. Ollama
Create a custom `Modelfile`:

```dockerfile
FROM ./aether_coder_q4_k_m.gguf

SYSTEM """You are Aether 2.5 Coder, an expert AI programming assistant."""

PARAMETER stop "<|im_start|>"
PARAMETER stop "<|im_end|>"