DavidAU/Qwen3.5-27B-Claude-4.6-OS-Auto-Variable-Thinking

VISIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Mar 6, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

DavidAU/Qwen3.5-27B-Claude-4.6-OS-Auto-Variable-Thinking is a 27 billion parameter Qwen 3.5 dense model fine-tuned by DavidAU using the Claude-4.6-OS dataset. This model features altered reasoning/thinking blocks and block sizes, aiming to improve performance without negatively impacting its strong benchmarks. It is a multimodal model with native vision capabilities and a default context length of 262,144 tokens, excelling in complex reasoning tasks and general agentic usage.

Loading preview...

Model Overview

This model, DavidAU/Qwen3.5-27B-Claude-4.6-OS-Auto-Variable-Thinking, is a 27 billion parameter variant of the Qwen 3.5 dense model. It has been fine-tuned by DavidAU using the Claude-4.6-OS dataset, specifically modifying its internal reasoning and thinking block structures and sizes. The fine-tuning process was designed to enhance these aspects while preserving the base model's strong benchmark performance.

Key Capabilities

  • Enhanced Reasoning: Features altered reasoning/thinking blocks for improved cognitive processes, often exhibiting Gemini Pro-like reasoning for complex tasks.
  • Multimodal Support: Fully supports vision (image) inputs and is built on the Qwen3.5 foundation, which includes unified vision-language capabilities.
  • Extended Context Window: Natively supports a context length of 262,144 tokens, extensible up to 1,010,000 tokens using YaRN scaling techniques.
  • Strong Benchmarks: Maintains high performance across various benchmarks, including knowledge, instruction following, long context, STEM & reasoning, coding, and general agent tasks.
  • Tool Calling: Excels in tool calling capabilities, recommended for use with Qwen-Agent and Qwen Code for building agent applications.

Good for

  • Complex Problem Solving: Ideal for tasks requiring deep reasoning and problem-solving, leveraging its optimized thinking blocks.
  • Multimodal Applications: Suitable for applications involving both text and image inputs, such as visual question answering and document understanding.
  • Agentic Workflows: Highly effective for developing AI agents that require tool use and interaction with external environments.
  • Long Context Processing: Excellent for tasks that demand processing and understanding of very long documents or conversations.

Finetunes
1 models