Cthulhu 7B v1.4 Overview

Cthulhu 7B v1.4 is a 7 billion parameter language model developed by EldritchLabs, based on the Mistral 7B v0.1 architecture. This model is a fully uncensored finetune, specifically trained on a compact dataset focused on Cthulhu and Goetia lore. It was trained for 3 epochs, achieving a loss of 0.1916 and a mean token accuracy of 0.9423.

Key Capabilities

• Uncensored Content Generation: Designed to produce narratives and roleplay content without censorship, including potentially violent and graphic erotic themes.
• Specialized Lore: Finetuned on Cthulhu and Goetia lore, making it adept at generating content within these specific domains.
• Performance: Achieves a Q0 Score of 8501 with Q6_K quantization, indicating strong performance for its size.
• Refusal-Free: Demonstrated 0 refusals out of 100 test cases, ensuring direct and unfiltered responses.

Good For

• Creative Writing: Ideal for generating dark fantasy, horror, or occult-themed stories and narratives.
• Roleplaying: Suitable for roleplaying scenarios that require uncensored and lore-specific interactions.
• Experimental Content: Users seeking to explore the boundaries of AI-generated content, particularly in mature or niche genres.

Note: Users are advised to adjust their system prompts accordingly and utilize the Mistral NonTekken chat template for optimal performance, especially given the model's uncensored nature.