Altworld/Astrea-R8-Chat-9B

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 20, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Altworld/Astrea-R8-Chat-9B is a 9 billion parameter model developed by Altworld, built upon the Qwen3.5-9B base model. It is specifically fine-tuned for creative writing, excelling at generating human-like prose, maintaining narrative consistency, and understanding scene context. This compact model outperforms larger creative writing specialists in quality and coherence, making it efficient for narrative generation tasks.

Loading preview...

Altworld Astrea-R8-Chat-9B: A Compact Model for Creative Writing

Altworld Astrea is a 9 billion parameter model, fine-tuned from the Qwen3.5-9B base, specifically designed for high-quality creative writing. It focuses on generating prose that sounds human, maintains narrative consistency, and understands scene context, often outperforming larger models in these areas.

Key Capabilities

  • Human-like Prose: Achieves a low "AI-sounding writing" score of 2.5, significantly less than rivals (8-19), indicating more natural language generation.
  • Narrative Coherence: Excels at maintaining story grounding, keeping 73 out of 100 scenes straight, compared to 44 for the strongest 12B prose model.
  • Contextual Understanding: Understands character motivations and plot details, leading to more cohesive and engaging narratives.
  • Efficiency: At 9B parameters, it is smaller than its creative writing rivals (typically 12B), allowing for faster inference and lower operational costs.

Good For

  • Creative Writing Applications: Ideal for generating stories, dialogues, and other narrative content where human-like quality and consistency are paramount.
  • Interactive Storytelling: Its ability to maintain grounding makes it suitable for applications requiring persistent narrative states.
  • Resource-Constrained Environments: Its compact size allows it to run efficiently on hardware where larger models might struggle.