DavidAU/L3.1-Dark-Reasoning-Unholy-Hermes-R1-Uncensored-8B is an 8 billion parameter, uncensored Llama 3.1-based model with a 32768-token context length, developed by DavidAU. It integrates variable control reasoning from NousResearch's DeepHermes model, optimized for creative use cases, logic puzzles, and riddles. This model is specifically designed to produce graphic horror and NSFW content, offering a "darkened" and compressed reasoning style that can be further customized via system prompts.
Loading preview...
L3.1-Dark-Reasoning-Unholy-Hermes-R1-Uncensored-8B Overview
This 8 billion parameter model, developed by DavidAU, is an uncensored variant based on the Llama 3.1 architecture, featuring a 32768-token context length. It incorporates variable control reasoning technology from NousResearch's DeepHermes model, offering a unique "darkened" and compressed reasoning style. The model is explicitly designed for NSFW, graphic horror, R/X-rated, and uncensored content generation, making it distinct from many other LLMs.
Key Capabilities
- Variable Control Reasoning: Leverages DeepHermes technology for enhanced problem-solving, logic puzzles, and riddles, with reasoning power estimated at 50-67% of the original DeepHermes model.
- Uncensored & Graphic Content: Specializes in generating explicit and graphic horror, R/X-rated, and general uncensored content.
- Creative Use Cases: Optimized for creative writing and scenarios, allowing for deeper, richer "thoughts" at higher temperatures.
- System Prompt Customization: Reasoning can be activated, deactivated, or varied through specific system prompts, including "evil" or "creative" personas (e.g., a Joker-like AI).
- Temperature Sensitivity: Performs best with higher temperatures (1.2+, 2.2+) for more expansive and interesting reasoning.
Good For
- Generating graphic horror and explicit content: Ideal for developers and creators requiring uncensored and detailed dark narratives.
- Creative writing and roleplay: Excels in scenarios demanding imaginative and unconstrained outputs, especially with customized reasoning personas.
- Solving logic puzzles and riddles: Utilizes its enhanced "thinking" systems to tackle problems beyond typical Llama 3.1 capabilities.
- Experimenting with variable reasoning: Users can fine-tune the model's internal thought processes via system prompts to achieve desired output styles.
Note: This model requires the Llama 3 Instruct template and benefits from medium to long prompts for optimal performance. Users should be aware that reasoning blocks may sometimes require regeneration to activate or produce the best results.