Overview

MS3.2-24B-Penumbra-Aether is a 24 billion parameter language model developed by Vortex5. It was created using a custom merging method, combining several base models: MS3.2-24B-Chaos-Skies, Cydonia-24B-v4.3, Hearthfire-24B, and ms3.2-24b-longform. The merge configuration utilized specific parameters for strength, flavor, paradox, cube dimensions, steps, and boost, and was processed using bfloat16 data type.

Key Capabilities

• Model Merging: Built from multiple specialized models to combine their strengths.
• Custom Merge Method: Utilizes a unique hpq merge method with fine-tuned parameters.
• Extended Context: Supports a context length of 32768 tokens, beneficial for longer interactions and narratives.

Good For

• Storytelling: Designed to generate coherent and engaging narratives.
• Roleplay: Optimized for interactive and character-driven conversational scenarios.
• Creative Writing: Excels in various forms of creative text generation, including prose and imaginative content.