---
license: apache-2.0
base_model: Ruler97/Godoter-27B
library_name: gguf
tags:
- godot
- godot-engine
- gdscript
- game-development
- code
- gguf
- llama-cpp
- qwen3
pipeline_tag: text-generation
---

# Godoter — GGUF quantizations

 **Unofficial hobby project — NOT made by Godot.** Godoter is a personal, non-commercial fine-tune made by an independent hobbyist for fun and learning. It is **not made, created, maintained, affiliated with, sponsored by, or endorsed by the Godot Foundation or the Godot Engine team**. *"Godot"* is a trademark of the Godot Foundation, used only to describe what the model is for. Base model: **Qwen3.6-27B** (Apache 2.0); training data includes the **Godot documentation (CC-BY 4.0)**.

GGUF builds of **[Ruler97/Godoter-27B](https://huggingface.co/Ruler97/Godoter-27B)**, a Qwen3.6-27B
fine-tune specialized in **Godot 4 / GDScript**. Ready to drop into **LM Studio**,
**llama.cpp**, **Ollama** or any GGUF runner.

## Which file to download

| File | Quant | Size | RAM/VRAM (rough) | When to use |
|------|-------|------|------------------|-------------|
| `Godoter-Q4_K_M.gguf` | Q4_K_M | ~16 GB | 16 GB+ | smallest & fastest, light quality loss |
| `Godoter-Q6_K.gguf` | Q6_K | ~21 GB | 24 GB+ | **recommended** — best quality/size balance |
| `Godoter-Q8_0.gguf` | Q8_0 | ~27 GB | 32 GB+ | highest fidelity, closest to the merged model |

If unsure, take **Q6_K**. Drop to **Q4_K_M** if you're tight on VRAM.

## Benchmark (Q6, vs base Qwen3.6-27B)

| Test | Base | Godoter |
|------|:----:|:-------:|
| **Easy tasks** (Godot 4 migration traps) | 80% | **100%** |
| **Hard tasks** (advanced Godot 4) | 68% | **96%** |

![Benchmark](godoter_benchmark.png)

The base model, on hard questions, often drifts to the wrong framework (HTML/JS, Python,
gRPC, Unity) or doesn't finish. Godoter stays anchored to Godot 4 GDScript and delivers
correct, complete code.

## How to run

**LM Studio** — put the file under `models/<publisher>/Godoter/` and load it from the UI.

**llama.cpp**
```bash
llama-server -m Godoter-Q6_K.gguf -c 8192 --jinja -ngl 99
# then open http://127.0.0.1:8080
```

**Ollama**
```bash
# Modelfile
echo 'FROM ./Godoter-Q6_K.gguf' > Modelfile
ollama create godoter -f Modelfile
ollama run godoter "In Godot 4, connect a Button's pressed signal in code."
```

## About the model

Fine-tune of `unsloth/Qwen3.6-27B` via QLoRA (r=16), trained on ~18,000 Godot 4 examples
grounded in the official docs, plus 666 complete multi-file Godot 4 systems. See the
[main model card](https://huggingface.co/Ruler97/Godoter-27B) for full details.

## License & attribution

Derivative of **Qwen3.6-27B** (Apache 2.0) — released under **Apache 2.0**. Built with Unsloth.