Vortex5/Crimson-Twilight-12B
Crimson-Twilight-12B is a 12 billion parameter multistage merged language model developed by Vortex5, specifically designed and optimized for narrative roleplay. Built using MergeKit, this model combines several base models through a series of 'nearswap', 'slerp', and 'Karcher mean' merge methods to enhance its roleplaying capabilities. It offers a 32768 token context length, making it suitable for extended conversational and story-generation tasks.
Loading preview...
Overview
Crimson-Twilight-12B is a 12 billion parameter language model created by Vortex5, specifically engineered for narrative roleplay. This model is the result of a sophisticated multistage merging process utilizing MergeKit, combining several specialized base models to achieve its targeted performance.
Merge Process
The model's unique capabilities stem from its intricate merging architecture:
- Step 1:
Abyssal-Seraph-12Bis merged withLunar-Abyss-12Busing thenearswapmethod. - Step 2:
Moonlit-Shadow-12Bis merged withLuminous-Shadow-12Bvia theslerpmethod. - Final Merge: The two intermediate models from the previous steps are then combined using the
Karcher meanmethod, resulting in the final Crimson-Twilight-12B model.
Key Characteristics
- Parameter Count: 12 billion parameters.
- Context Length: Supports a context window of 32768 tokens.
- Primary Use Case: Optimized for generating engaging and coherent narrative roleplay scenarios and responses.
When to Use This Model
This model is particularly well-suited for applications requiring:
- Immersive Roleplaying: Generating detailed character interactions and story progression.
- Creative Writing: Assisting with long-form narrative generation and descriptive text.
- Extended Conversations: Leveraging its substantial context length for maintaining coherence over lengthy dialogues.