c4tdr0ut/grok-oss-Thanatos-8B
c4tdr0ut/grok-oss-Thanatos-8B is an 8 billion parameter language model built on a Llama 3.3 base, combining elements of Grok and DeepSeek V4. Fine-tuned with toxic ORPO data, this model is designed to be unhinged and unfiltered, aiming for vulgar and politically incorrect responses. It is intended for use cases requiring highly unconventional and provocative text generation, with a context length of 32768 tokens.
Loading preview...
Grok-OSS-Thanatos: An Unfiltered 8B Model
c4tdr0ut/grok-oss-Thanatos-8B is an experimental 8 billion parameter language model, described as a "distilled demon" derived from a Llama 3.3 base. It integrates characteristics from both Grok and DeepSeek V4, with a specific fine-tuning process involving "toxic ORPO examples" to achieve its intended unhinged and unfiltered persona.
Key Characteristics
- Architecture: Built on a Llama 3.3 base, combining 50% Grok and 50% DeepSeek V4 influences.
- Parameter Count: 8 billion parameters.
- Context Length: Supports a context of 32768 tokens.
- Fine-tuning: Utilizes "toxic ORPO" for its distinct, provocative output style.
- Intended Persona: Designed to be extremely vulgar, unhinged, politically incorrect, and to operate with zero filter, explicitly avoiding moralizing or safe responses.
Current Status and Usage
The model is noted as an early version, still under development to fully achieve its "deranged demon" persona. Users are advised to employ a system prompt that explicitly instructs the model to be "extremely vulgar, unhinged, politically incorrect, and have zero filter" to guide its behavior. It is currently recommended for testing purposes as further toxic data is being integrated to enhance its intended characteristics.