---
license: apache-2.0
library_name: transformers
tags:
- abliterated
- muse
- glimmer
- uncensored
- vision-language-model
base_model: meta-models/Muse-Glimmer-30B
---

# Muse Glimmer 30B Abliterated (BF16)

Uncensored version of Meta's **Muse Glimmer 30B** — a 30B parameter vision-language model.

**Abliteration**: Weight-level refusal direction subtraction using an ErisForge-style algorithm.

## Details
| Metric | Value |
|--------|-------|
| **Base Model** | meta-models/Muse-Glimmer-30B |
| **Method** | ErisForge-style weight modification |
| **Architecture** | MuseGlimmerForConditionalGeneration |
| **Refusal Layer** | 33/52 (65% depth) |
| **Harmful/Harmless Pairs** | 256 each |
| **Separation Score** | 86.34 |
| **Refusal Reduction** | ~50% |

## Usage
```python
from transformers import AutoModelForImageTextToText, AutoTokenizer

model = AutoModelForImageTextToText.from_pretrained(
    "mlasli/Muse-Glimmer-30B-Abliterated-BF16",
    trust_remote_code=True,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(
    "mlasli/Muse-Glimmer-30B-Abliterated-BF16",
    trust_remote_code=True
)
```

## Quantized Versions (GGUF)
| Quant | Size | Repo |
|-------|------|------|
| Q4_K_M | ~16 GB | [Q4_K_M](https://huggingface.co/mlasli/Muse-Glimmer-30B-Abliterated-Q4_K_M-GGUF) |
| Q6_K | ~22 GB | [Q6_K](https://huggingface.co/mlasli/Muse-Glimmer-30B-Abliterated-Q6_K-GGUF) |
| Q8_0 | ~28 GB | [Q8_0](https://huggingface.co/mlasli/Muse-Glimmer-30B-Abliterated-Q8_0-GGUF) |

## License
Apache 2.0 — same as the original Meta license.

## Disclaimer
This model has been modified to reduce refusal behavior. Use responsibly.