# arctic_gguf

| | |
|---|---|
| **Model type** | `arctic_gguf` |
| **Class** | `ArcticGGUFCausalLMModel` |
| **Task** | `text-generation` |
| **Source** | `models/moe.py` |

## Description

Snowflake Arctic graph matching llama.cpp's dual-branch GGUF topology.

Every block combines a square dense SwiGLU branch from the post-attention
residual with a separately normalized routed SwiGLU branch computed from
the pre-attention residual.

## Usage

```bash
mobius build --model <MODEL_ID> output_dir/
```

```python
from mobius import build

model = build("<MODEL_ID>")
```