empero-ai/Qwable-9B-Claude-Fable-5

Hugging Face
VISIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Jun 15, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Qwable-9B-Claude-Fable-5 by Empero is a 9 billion parameter, full-parameter supervised fine-tune of Qwen3.5-9B, a multimodal model with a hybrid attention stack. It is specifically optimized for agentic coding and reasoning tasks, distilled from Claude Fable 5 and GPT-5.5 terminal agent traces. This model excels at imitating complex reasoning and tool-use styles on long, multi-turn coding and agent tasks, making it a strong choice for developers needing advanced code generation and problem-solving capabilities.

Loading preview...

Overview

Qwable-9B-Claude-Fable-5 is a 9 billion parameter model developed by Empero, built upon the Qwen3.5-9B base. It's a full-parameter supervised fine-tune, specifically designed to excel in agentic coding and reasoning tasks. The model learns by imitating the reasoning and tool-use styles of advanced assistants like Claude Fable 5 and a GPT-5.5 terminal agent, using a curated mix of their traces.

Key Capabilities

  • Advanced Coding & Agentic Behavior: Demonstrates strong performance in generating correct and idiomatic code, handling terminal/agent tasks, and applying security-aware judgment.
  • Reasoning-First Approach: Each response begins with a <think> block, providing a transparent reasoning process before the final answer.
  • Long Context Handling: Trained with full-length traces up to 76,800 tokens, enabling it to manage long, multi-turn conversations without truncation.
  • Text-Only Fine-tune: While based on a multimodal architecture, the fine-tuning was text-only, focusing its capabilities on language-based tasks.

Good For

  • Complex Code Generation: Ideal for developers requiring sophisticated Python functions, script generation, and understanding of current tooling.
  • Agentic Task Automation: Suitable for tasks involving multi-step problem-solving, terminal interactions, and tool-use simulation.
  • Reasoning-Intensive Applications: Excellent for scenarios where a clear, step-by-step reasoning process is beneficial, such as debugging or complex problem analysis.
  • Research & Experimentation: Released under Apache-2.0, it's a valuable resource for exploring advanced fine-tuning techniques and agentic AI behaviors.