AuriAetherwiing/TQ3.5-27B-Musica-v1
AuriAetherwiing/TQ3.5-27B-Musica-v1 is a 27 billion parameter Qwen3.5-based language model fine-tuned for roleplay, story generation, and conversational tasks, supporting a 32768 token context length. It is designed for steerable output and reduced refusal rates, offering both reasoning and non-reasoning modes with multiple reasoning styles. This model excels in creative writing and interactive dialogue applications.
Loading preview...
AuriAetherwiing/TQ3.5-27B-Musica-v1 Overview
This model is a 27 billion parameter variant of Qwen3.5, specifically fine-tuned by AuriAetherwiing for enhanced roleplay, story generation, and conversational capabilities. It builds upon ArliAI's derestricted Qwen3.5-27B base, aiming to provide more steerable outputs and a reduced refusal rate.
Key Capabilities
- Roleplay and Story Generation: Optimized for creative writing and generating engaging narratives.
- Conversational Fluency: Designed for natural and interactive dialogue.
- Flexible Reasoning: Supports both general conversational modes and dedicated reasoning modes, including several distinct reasoning styles that can be evoked through rerolling.
- Reduced Refusal Rate: Aims to provide more compliant and less restrictive responses compared to its base model.
- Context Length: Utilizes a substantial 32768 token context window, beneficial for longer interactions and complex narratives.
Training Details
The model was trained using a Rank 64, alpha 64 LoRA adapter for two epochs on top of ArliAI's Derestricted Qwen3.5-27B. The training was sponsored by ArliAI and took approximately 17 hours on OwenArli's 2xRTX Pro 6000 Blackwell setup.
Recommended Samplers
For optimal performance, the developers recommend using a Temperature of 1, NSigma of 2, and Min-P of 0.02.
Good For
- Applications requiring creative text generation, such as interactive fiction or scriptwriting.
- Chatbots and virtual assistants where nuanced, steerable conversations are important.
- Scenarios demanding flexible reasoning capabilities within a conversational context.