RicardoEstep/RPBizkit-v7-12B

RicardoEstep/RPBizkit-v7-12B: A Specialized Mistral Nemo Merge

RicardoEstep/RPBizkit-v7-12B is a 12.2 billion parameter experimental language model built upon the Mistral Nemo architecture. It leverages a sophisticated five-part "Model Stock" merging process, utilizing Mergekit to combine and filter various base models. This approach aims to create a specialized model by cleanly mixing and filtering noise from each component.

Key Capabilities & Design:

• Multi-Stage Merging: The model is constructed through five distinct merging phases, each focusing on a specific aspect:
  • Intelligence Core: Integrates models like Mistral-Nemo-Instruct-2407-abliterated and Mistral-Nemo-2407-12B-Thinking-Claude-Gemini-GPT5.2-Uncensored-HERETIC to enhance general intelligence.
  • RP & Characterization Core: Incorporates models such as Krix-12B-Model_Stock and Himeyuri-v0.1-12B to specialize in roleplay and character depth.
  • Narrative Core: Merges models like Ohashi-NeMo-12B and MN-12b-RP-Ink to improve storytelling and narrative flow.
  • Substance Core: Combines models including Nera_Noctis-12B and Impish_Bloodmoon_12B for content richness.
  • Final Mix: A final merge of the four specialized cores.
• Extended Context: The model is designed to support a full 128K context size, with clean tokenizer and embedding sizes (131072).
• Alpaca Chat Template: Recommended for use with the "Alpaca (with RAW inputs)" chat template, though it also supports an optional optional_chat_template.jinja.

Ideal Use Cases:

• Roleplay (RP) Scenarios: Excels in generating detailed and engaging character interactions and roleplay narratives.
• Creative Writing: Suitable for tasks requiring rich storytelling, character development, and nuanced dialogue.
• Uncensored Content Generation: Explicitly designed with "RP Uncensored" models, making it suitable for less restricted content generation, with the creator noting no responsibility for its use.