lemuralabs/Qwen3.6-27B-Fable-Agentic

VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 23, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Qwen3.6-27B-Fable-Agentic by lemuralabs is a 27 billion parameter agentic model fine-tuned from Jackrong/Qwopus3.6-27B-v2. It excels in step-by-step reasoning and multi-turn tool-calling, while maintaining top-tier general knowledge. This model is optimized for complex tasks requiring both advanced reasoning and external tool interaction.

Loading preview...

Model Overview

Qwen3.6-27B-Fable-Agentic is a 27 billion parameter model developed by lemuralabs, fine-tuned from the Jackrong/Qwopus3.6-27B-v2 base. Its core strength lies in its agentic capabilities, enabling it to perform multi-turn tool-calling and engage in step-by-step reasoning. The model demonstrates strong general knowledge and mathematical proficiency, making it suitable for complex problem-solving.

Key Capabilities

  • Agentic Tool-Calling: Designed for multi-turn interactions with external tools.
  • Chain-of-Thought Reasoning: Facilitates detailed, step-by-step problem-solving.
  • Top-Tier Knowledge: Achieves 88.29% on MMLU-Pro, showing improved knowledge over its base model.
  • Strong Math Performance: Scores 97.33% on GSM8K, indicating robust mathematical reasoning.
  • Multimodal Architecture: Supports both vision and text inputs.

Performance Highlights

Evaluated with an 8192-token budget to ensure full reasoning, the model shows competitive performance:

  • MMLU-Pro: 88.29% (outperforming its base model's 86.86%)
  • GSM8K: 97.33% (comparable to its base model's 98.00%)
  • GPQA-Diamond: 57.07% (comparable to its base model's 57.58%)

This indicates that the model preserves hard reasoning capabilities while enhancing general knowledge and integrating agentic features.