xtools-at/Qwen3.5-4B-Claude-3xMerge-Jackrong-UNCENSORED-heretic

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Oct 14, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The xtools-at/Qwen3.5-4B-Claude-3xMerge-Jackrong-UNCENSORED-heretic is a 4.5 billion parameter language model based on the Qwen3.5-4B architecture, created by xtools-at. It is a decensored version of a weighted merge of three Qwen3.5-4B fine-tunes by Jackrong, focusing on Claude Opus-style reasoning, coding, and structured multi-step problem-solving. This model is optimized for general reasoning, programming, debugging, and agent-oriented workflows, offering enhanced capabilities in complex problem decomposition and explicit reasoning structures.

Loading preview...

Overview

This model, xtools-at/Qwen3.5-4B-Claude-3xMerge-Jackrong-UNCENSORED-heretic, is a 4.5 billion parameter language model built upon the Qwen3.5-4B base architecture. It is a decensored version, created using Heretic v2.0.0.dev0, of a weighted merge of three fine-tuned Qwen3.5-4B models by Jackrong. The merge combines Claude Opus-style reasoning distillation, coding-oriented fine-tuning, and structured multi-step reasoning to create a compact, reasoning-focused checkpoint.

Key Capabilities

  • Enhanced Reasoning: Integrates general reasoning, analytical problem-solving, and instruction-following behaviors distilled from Claude Opus 4.6-style traces.
  • Coding and Debugging: Optimized for software development workflows, including code generation, debugging, and tool-assisted problem-solving through trace inversion and agent trajectories.
  • Structured Problem-Solving: Emphasizes explicit multi-step reasoning, complex problem decomposition, and consistent reasoning across tasks.
  • Decensored Output: Modified to reduce refusals, with a reported 4/100 refusals compared to 98/100 in the original merged model.

Good for

  • General Reasoning Tasks: Ideal for multi-step questions, logical deduction, and analytical problem-solving.
  • Programming and Development: Suitable for code generation, explanation, and debugging in various programming contexts.
  • Agentic Workflows: Experimentation with structured tool calls and iterative execution in compatible environments.
  • Local Inference: Deployment and evaluation in environments supporting the Qwen3.5 architecture, particularly where a compact, reasoning-oriented model is desired.

Finetunes
1 models