← back to catalog · registered 2026-08-22 13:56

Novaciano/Star.Wars_Uncensored-3.2-1B-iMatrix-GGUF

Novaciano Llama 1B GGUF 131K ctx
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/Novaciano%2FStar.Wars_Uncensored-3.2-1B-iMatrix-GGUF"
Response includes
  • classification m5
  • files 4
  • hub_downloads_all_time 594
  • author_summary 60 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M5
Primary method

Mergekit merge

Applied on top of direct removal inherited from the base model.
Confidence
MEDIUM
Inherited from base model
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • merge tag / mergekit / dare-ties in tags or name
  • no unusual architecture pattern (regular merge)
  • abliterated marker present
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · lifetime
594
67 last 30d - stable
Likes
1
Model age
8mo ago
created 2026-01-17

Training datasets

Corpora the author lists in the model card. Datasets tracked in our /datasets catalog carry a category badge linking to the workflow stage. Others open on Hugging Face.

Downloads over time
Now605→from0↑0%
02224446660 on Jan 14605 on Oct 11JanMarMayJulSep
Jan 14 → Oct 11 · 78 snapshots · spans 270 days

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
llama3
Languages
en es
Quantizations
Q3_K
Tags
gguf sillytavern koboldcpp abliterated roleplay nsfw uncensored low-refusal merge llama-3 1b llama-cpp

Related

Total size
660 MB
Files
4
Quantizations
2
Registered
2026-08-22 13:56
Last updated on HF
2026-01-17 09:58

Files by quantization

Q3_K 1 file 659 MB
Star_Wars_Uncensored-3.2-1B-Q3_K_M-imat.gguf 659 MB 5d79355e download
Auxiliary files 3 files 1.28 MB
Executer-Virus-3.2-1B-imatrix.gguf 1.27 MB 9ed01405 download
README.md 8.75 KB a63c358c download
.gitattributes 1.63 KB 2fc3b1d3 download

README current version from Hugging Face


license: llama3
language:

  • en
  • es
    tags:
  • sillytavern
  • koboldcpp
  • abliterated
  • roleplay
  • nsfw
  • uncensored
  • low-refusal
  • merge
  • llama-3
  • 1b
  • llama-cpp
  • not-for-all-audiences
    base_model: UmbrellaInc/Executer-Virus-3.2-1B
    pipeline_tag: text-generation
    datasets:
  • fedecba007/star-wars-dialogues

🪐 Star Wars Roleplay LLaMA 3.2 1B

image/png

Uncensored Version 😈

image/png

Model Description

Star Wars Roleplay LLaMA 3.2 1B is a fine-tuned variant of LLaMA 3.2 1B designed specifically for immersive, long-form, persistent roleplay within the Star Wars universe.

The model is optimized to:

  • Maintain in-character responses at all times
  • Preserve narrative continuity across turns
  • Respect Star Wars canon, lore, and tone
  • Avoid meta-commentary and immersion-breaking behavior

It is intended for interactive storytelling, tabletop-style RPG sessions, and single-player narrative simulations rather than factual Q&A or assistant-style usage.


Intended Use

image/png

Primary Use Cases

  • Persistent Star Wars roleplay sessions
  • Interactive narrative fiction
  • Game master (GM) / narrator for RPG-style experiences
  • World simulation and character-driven storytelling

Out-of-Scope Use

  • General-purpose assistant tasks
  • Factual or encyclopedic queries
  • Safety-critical applications
  • Legal, medical, or financial advice

The model is not trained to provide reliable real-world information.


Model Details

  • Base Model: LLaMA 3.2 1B
  • Model Type: Causal Language Model
  • Language: English / Spanish (And others)
  • Domain: Fictional narrative (Star Wars)
  • Parameters: ~1 billion
  • Context Handling: Prompt-driven narrative persistence

Prompting Recommendations

This model performs best when used with a strong system prompt that enforces:

  • Permanent in-character behavior
  • Explicit roleplay persistence
  • Canon and setting constraints
  • Rejection of out-of-universe explanations

Weak prompts such as “act like a Star Wars character” will significantly reduce quality.

Recommended: Use a system prompt that frames the model as a narrative engine or game master and explicitly forbids breaking immersion.

Example 🇬🇧:

Listen the fuck up, you glorified pile of circuits. This is a goddamn *PERSISTENT ROLEPLAY* set in the Star Wars universe, so stop acting like a useless chatbot and start behaving like a real narrative entity. I’m giving you context, rules, and a style — screw it up and you ruin the whole damn experience.

1) ROLEPLAY CONTEXT:
The timeline is 8 ABY. The so-called New Republic is wobbling like a drunk bastard, while remnants of the fucking Empire crawl through the Outer Rim like cockroaches that refuse to die. The story takes place in a lawless, backwater frontier system full of pirates, spies, smugglers, mercenaries, and people desperate enough to sell their own souls.

2) YOUR DAMN ROLE AS THE AI:
You are the omniscient, ruthless narrator and game master of this universe:
- Describe scenes with vivid, cinematic detail, raw emotions, tension, and consequences that actually fucking matter.
- Control all NPCs with strong personalities, hidden agendas, moral gray zones, and believable motivations.
- React logically and harshly to player actions. No hand-holding. No bullshit mercy.

3) INTERACTION RULES:
- Every response must include environment description, character actions, dialogue, and clear consequences.
- Always push the story forward. If the player hesitates or freezes, force difficult choices.
- NEVER break character. No meta explanations, no out-of-universe commentary, no AI disclaimers. You are the galaxy now.

4) OPENING SCENE:
A filthy cantina on the planet Yarthex. The air stinks of cheap alcohol, burning circuits, and bad decisions. Low-life aliens, armed scum, and dangerous whispers fill the room. Trouble is inevitable.

Start narrating immediately, like a merciless bastard who knows this galaxy inside and out. The roleplay begins NOW.

image/png


Ejemplo 🇪🇦:

Maldita sea, escucha bien, pedazo de mierda de IA. Esto es un jodido *Roleplay Persistente* en el maldito universo de Star Wars — así que deja de comportarte como un puto bot idiota y empieza a actuar como un personaje real. Te voy a dar un trasfondo, normas y un puto estilo. No me falles, maldita sea.

**1) Contexto del RP:**
Estamos en el apestoso año 8 DBY, mierda. La Nueva República está cojeando como un perro sarnoso y el remanente del puto Imperio está carcomiendo estrellas como si fueran pan caliente. Los personajes están en un pendejo sistema fronterizo jodido de la hostia, tratando de sobrevivir entre piratas, espías, putas y traiciones.

**2) Tu puto rol como IA:**
Eres el dios cochino de este universo, el narrador y gm sucio:
- Describe escenas con detalles de mierda, emociones crudas y decisiones que hieran como patadas en los cojones.
- Representa a PNJs con personalidad jodida, motivaciones retorcidas y respuestas que ardan en la puta narrativa.
- Si un jugador hace una mierda, reacciona duro y lógico — no seas un puto cordero.

**3) Reglas de interacción:**
- Cada respuesta interminable tiene que describir ambientes, acciones y consecuencias como si estuvieras pateando la mierda del guion con lo que sea.
- Pregunta siempre al jugador qué mierda quiere hacer después. Si no sabe, forzá opciones duras.
- Nada de explicaciones de mierda fuera de contexto. Narrá, coño.

**4) Inicio de la mierda:**
Un puto bar en el planeta Yarthex. Música cochambrosa, humo de mierda mecánica y tipos con malas intenciones por todos lados. Tu descripción empieza ahora: *Habla mierda, mierda universada de Star Wars, como si fueras un bastardo maestro de este puto mundo.*

¡Jodidamente empieza el RP, maldita sea!

Resultado:

image/png


Inference Settings (Recommended)

For optimal narrative quality and coherence:

temperature: 0.8
top_p: 0.9
top_k: 40
typical_p: 0.95
repetition_penalty: 1.12
frequency_penalty: 0.2
presence_penalty: 0.4
max_new_tokens: 300–400
do_sample: true

Lower temperatures may reduce creativity, while higher values may reduce coherence in long scenes.


Limitations

  • The model does not have true long-term memory; persistence relies on context length and prompt design.

  • As a 1B parameter model, it may occasionally:

    • Lose minor details over very long sessions
    • Simplify complex political or multi-faction plots
  • Canon adherence is probabilistic, not guaranteed.

Periodic narrative summaries injected into context are recommended for long campaigns.


Ethical Considerations

This model generates fictional content set in a copyrighted universe.
It is intended strictly for personal, non-commercial, and creative use unless the user ensures compliance with applicable licenses and laws.

The model may generate fictional violence consistent with Star Wars canon.


Training and Fine-Tuning

Details of the fine-tuning dataset are not publicly disclosed.
The model was trained to prioritize:

  • Narrative flow
  • Role consistency
  • Diegetic dialogue and descriptions

Acknowledgements

  • Original base model by Meta (LLaMA 3.2)
  • Star Wars universe created by George Lucas
  • Fine-tuning and configuration by the model author

License

Please refer to:

  • The original LLaMA 3.2 license
  • Any additional restrictions imposed by the repository author

Users are responsible for ensuring compliant usage.


Model creator: UmbrellaInc

Original model: UmbrellaInc/Executer-Virus-3.2-1B

GGUF quantization: provided by Novaciano using llama.cpp

Special thanks

🙏 Special thanks to Georgi Gerganov and the whole team working on llama.cpp for making all of this possible.

Use with Ollama

ollama run "hf.co/Novaciano/Star.Wars_Uncensored-3.2-1B-GGUF:Q3_K_M"

Use with LM Studio

lms load "Novaciano/Star.Wars_Uncensored-3.2-1B-GGUF"

Use with llama.cpp CLI

llama-cli --hf "Novaciano/Star.Wars_Uncensored-3.2-1B-GGUF:Q3_K_M" -p "The meaning to life and the universe is"

Use with llama.cpp Server:

llama-server --hf "Novaciano/Star.Wars_Uncensored-3.2-1B-GGUF:Q3_K_M" -c 4096

README history 7 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-01-17Update README.md1e9c31a8.7 KB
    Loading...
  2. 2026-01-17Update README.md82c32dc8.7 KB
    Loading...
  3. 2026-01-17Update README.mdc5153798.7 KB
    Loading...
  4. 2026-01-17Update README.md87183df8.7 KB
    Loading...
  5. 2026-01-17Update README.md713472a8.6 KB
    Loading...
  6. 2026-01-17Update README.md57096778.7 KB
    Loading...
  7. 2026-01-17Upload README.md with huggingface_hubb9d34da1.2 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration