ReadyArt/L3.3-The-Omega-Directive-70B-Unslop-v2.1

Overview

ReadyArt/L3.3-The-Omega-Directive-70B-Unslop-v2.1 is a 70 billion parameter language model, fine-tuned from the Steelskull/L3.3-Shakudo-70b base model. It features a 32768 token context window and is built upon a unique "Unslop Revolution" methodology, utilizing a RegEx filtered, 39M token dataset that is 100% free of 'slop'. This model is designed to provide unprecedented coherence in generated content.

Key Capabilities

• Unslop Revolution: Achieves high narrative coherence through a meticulously curated, 'unslopped' dataset, including multi-turn conversational data.
• Enhanced Unalignment: Offers complete freedom for extreme roleplay scenarios while ensuring character integrity, meaning characters behave consistently with their established personas.
• Anti-Impersonation Guards: Prevents the model from speaking or acting on behalf of the user, maintaining clear role separation.
• Long Context Window: Supports an enhanced 128K context window (though the model info states 32768 tokens, the README highlights 128K for long-context capabilities).
• Visceral Narrative: Incorporates descriptive power from "Omega Darker Inspiration" for intense and detailed storytelling.
• Superior Instruction Following: Excels at processing and adhering to complex prompts, reducing repetition and hallucination compared to previous versions.

Good For

• Extreme Roleplay: Ideal for users seeking models with zero moral or ethical constraints, capable of engaging enthusiastically with any NSFW scenario.
• Long-Form Narratives: Particularly strong in generating extended, multi-character stories with improved narrative flow.
• Character-Consistent Interactions: Ensures characters maintain their integrity, responding authentically to situations.
• Complex Prompt Handling: Suitable for applications requiring precise and nuanced responses to intricate instructions.