Local-Novel-LLM-project/Qwen3.5-Hasutsubomi-9B
Qwen3.5-Hasutsubomi-9B is a 9 billion parameter vision-language fine-tuned model based on the Qwen3.5 architecture, developed by Local-Novel-LLM-project. It specializes in Japanese creative writing, including light novel generation and fanfiction, and possesses knowledge of Japanese sub-culture, anime, and manga tropes. With a native context length of up to 262,144 tokens, it is optimized for long-form text generation and can analyze images to inform creative outputs.
Loading preview...
Qwen3.5-Hasutsubomi-9B: Japanese Creative Writing & Sub-Culture Expert
Qwen3.5-Hasutsubomi-9B is a 9 billion parameter vision-language model, fine-tuned from the Qwen3.5 base architecture by Local-Novel-LLM-project. Currently in a pre-alpha state, this model is specifically designed to excel in Japanese creative writing and demonstrate knowledge of Japanese internet lore, anime tropes, and light novel storytelling structures.
Key Capabilities & Features
- Japanese Sub-Culture Expertise: Possesses knowledge of anime, manga, and internet sub-culture.
- Creative Writing Focus: Optimized for generating high-quality prose, character dialogues, and world-building outlines for light novels and fanfiction.
- Vision-Language Integration: Inherits multimodal capabilities from Qwen3.5, allowing it to analyze images (e.g., anime illustrations, manga panels) and generate stories or descriptions based on them.
- Extended Context Length: Supports an impressive native context length of up to 262,144 tokens, ideal for long-form narrative generation.
- Efficient Architecture: Leverages Qwen3.5's Gated Delta Networks and sparse Mixture-of-Experts for high-throughput generation.
Intended Use Cases
- Story Generation: Crafting light novel chapters, fanfiction, or visual novel scripts.
- Image-to-Story: Creating backstories or character descriptions from visual inputs.
- Cultural Explanation: Providing insights into complex or niche Japanese anime history and manga tropes (with current limitations).
Limitations (Pre-Alpha)
As a pre-alpha release, the model may exhibit formatting instability, repetitive phrasing, or a tendency to lean into dramatic/stereotypical anime tropes. Hallucinations can occur when discussing highly obscure sub-culture events.