Pranavz/qwen-4b-2507-rp-mahou
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Apr 29, 2026License:apache-2.0Architecture:Transformer Open Weights Cold
Pranavz/qwen-4b-2507-rp-mahou is a 4 billion parameter language model, fine-tuned from Qwen/Qwen3-4B-Instruct-2507 by Pranavz. It specializes in creative roleplay and character interaction, trained on the flame-kindling-v1 dataset. This model excels at generating vivid, action-oriented narrative responses for immersive roleplaying scenarios, supporting a 32768 token context length.
Loading preview...
Model Overview
Pranavz/qwen-4b-2507-rp-mahou is a 4 billion parameter language model, derived from the Qwen3-4B-Instruct-2507 base model developed by the Qwen team. This model has undergone full-parameter supervised fine-tuning (SFT) on the flammenai/flame-kindling-v1 dataset, specifically curated for creative roleplay and character interaction.
Key Capabilities
- Creative Roleplay: Optimized for generating vivid, in-character responses, often utilizing asterisks for actions, as seen in its training data.
- Full-Sequence SFT: Utilizes full-sequence SFT rather than LoRA, indicating a comprehensive fine-tuning approach.
- Extended Context: Supports a context length of 32768 tokens, allowing for longer and more complex roleplay scenarios.
- Qwen3 Chat Template: Employs the Qwen3 chat template, with a specific recommendation to use
enable_thinking=Falsefor roleplay to prevent Chain-of-Thought reasoning.
Recommended Use Cases
- Immersive Roleplaying: Ideal for applications requiring detailed and imaginative character interactions.
- Creative Writing Assistance: Can be used to generate narrative segments or character dialogues in a vivid style.
Limitations
- Specific Tone: The model's output reflects the particular tone and style of its training dataset (vivid, action-asterisk style).
- Safety: Safety tuning is limited to what the base Qwen model provides.
- Language: Primarily supports English language generation.