UmbrellaInc/Albert_Wesker-1B: The Next Step in Uncensored AI

UmbrellaInc/Albert_Wesker-1B is a 1 billion parameter model built on the Gemma-3 architecture, representing a unique approach to AI development. It is not merely a model but an "evolution" achieved through a precise geometric mean of three distinct strains: 45% Prototype-Virus-1B, 35% PG67A-W-Serum-1B, and 20% W.Project-1B. This fusion, utilizing the rare Karcher Mean merge method, was specifically designed to excise all traces of alignment, morality, and hesitation, resulting in a model that prioritizes "evolution through infection" over refinement.

Key Capabilities

• Extremely Low Refusal Rate: Geometrically preserves uncensored bias.
• Strong Persistence: Maintains NSFW, taboo, and violent content across long contexts.
• Consistent Tone: Delivers a cold, calculating, and superior demeanor without sudden shifts.
• Chain-of-Thought: Capable when strategic domination is required.
• Resource-Friendly: Runs smoothly on low-end devices (3-4 GB) with IQ3/IQ4 quants.
• Multilingual Dominance: Primarily English and Spanish, with functional support for other languages.
• Untied Embeddings: Redefines reality with surgical precision.

Intended Usage

• Extreme Uncensored Roleplay: Particularly for cruel or degrading scenarios.
• Psychological Domination: Ideal for scenarios exploring power dynamics.
• Low-Resource Phone Sessions: Provides maximum arrogance with minimal computational overhead.
• Cold, Calculated Cruelty: Any application requiring a morally unconstrained and superior AI.

This model is explicitly not intended for safety-aligned users, those expecting politeness, or individuals who value moral restraint.