wayzer2321/Luna-27B-GGUF-Archive overview
Luna 3.8 27B GGUF — archived edition / прошлая версия ARCHIVED / ПРОШЛАЯ ВЕРСИЯ. This repository preserves an older Luna release. Актуальная версия / Current r…
Runs locally from ~12.57 GB disk (16 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| luna3.8-27b-q3.gguf | GGUF | Q3 | 12.57 GB | Download |
Model Details
| Model ID | wayzer2321/Luna-27B-GGUF-Archive |
|---|---|
| Author | wayzer2321 |
| Pipeline | text-generation |
| License | apache-2.0 |
| Base model | Qwen/Qwen3.8-27B |
| Last modified | 2026-10-06T13:12:44.000Z |
Model README
---
language:
- en
- ru
license: apache-2.0
tags:
- legacy
- archived
- gguf
- qwen
- qwen3.8
- 27b
- heavyweight
- character
- persona
- troll
- trolling
- chaos
- unhinged
- post-irony
- dark-humor
- conversational
- fine-tune
- text-generation
base_model: Qwen/Qwen3.8-27B
pipeline_tag: text-generation
model_name: Luna 3.8 27B GGUF (Archived)
---
Luna 3.8 27B GGUF — archived edition / прошлая версия
> ARCHIVED / ПРОШЛАЯ ВЕРСИЯ. This repository preserves an older Luna release.
> Актуальная версия / Current release: Luna v2 · 9B GGUF.
> Все релизы / Release catalog: luna-all-versions.
Release status
This is a separate older 27B checkpoint. It is not a 27B edition of the current Luna v2 checkpoint.
Этот репозиторий оставлен для старых загрузок и сравнения версий.
Для знакомства с актуальной Луной используйте Luna v2, ссылка выше.
Название базовой модели и число параметров не обозначают порядок релизов Луны.
Archived model details
| Property | Value |
| :--- | :--- |
| Status | Archived / previous release |
| Checkpoint | Older separate Luna 27B edition |
| Base model | Qwen/Qwen3.8-27B |
| File | luna3.8-27b-q3.gguf |
| Quantization | Q3_K_M |
| Format | GGUF |
Luna is Wayzer's conversational character: informal slang, sarcasm,
surreal humor, and creator lore. Responses from this archived checkpoint
are not representative of the current Luna v2 release.
Historical sampling settings
These are the settings previously published for this archived model:
- Temperature:
0.75–0.88 - Repeat penalty:
1.08–1.10 - Top-p:
0.92 - Context length:
2048
Memory use and generation speed depend on hardware, context, and backend.
Local import (Ollama)
For this archived file only:
FROM ./luna3.8-27b-q3.gguf
Save the line above in Modelfile next to the GGUF, then run:
ollama create luna-27b-archive -f Modelfile
ollama run luna-27b-archive
Historical dialogue archive
Earlier dialogue examples and release catalog.
These examples are historical and are not a benchmark of Luna v2.
Run wayzer2321/Luna-27B-GGUF-Archive with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models