xtools-at/Qwen3.5-4B-MultiDistill-3xMerge

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Oct 12, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

xtools-at/Qwen3.5-4B-MultiDistill-3xMerge is a 4.5 billion parameter experimental checkpoint merge based on the Qwen3.5-4B architecture, designed for enhanced reasoning capabilities. It combines three distinct fine-tunes: Claude Opus 4.6-style reasoning distillation, Empero's Qwen3.8 teacher distillation, and DeepSeek-V4-Pro mathematics and STEM reasoning. This model aims to provide strong general reasoning, efficient problem-solving, and specialized mathematical and scientific analytical skills within a compact form factor, making it suitable for complex logical and quantitative tasks.

Loading preview...

Qwen3.5-4B Triple-Distill Reasoning Merge

This model, xtools-at/Qwen3.5-4B-MultiDistill-3xMerge, is an experimental 4.5 billion parameter checkpoint merge built upon the Qwen3.5-4B base architecture. It uniquely combines three distinct reasoning-focused fine-tunes to create a compact yet powerful model for analytical tasks.

Key Capabilities

  • Enhanced General Reasoning: Integrates structured, efficient reasoning patterns distilled from Claude Opus 4.6-style traces, focusing on logical deduction and problem-solving.
  • Broad Instruction Following: Incorporates teacher-generated reasoning traces from Empero's Qwen3.8 distillation, covering general knowledge and diverse instruction-following scenarios.
  • Specialized Mathematics & STEM: Leverages distillation from DeepSeek-V4-Pro, providing strong capabilities in detailed mathematical derivations, scientific problem-solving, and multi-step quantitative reasoning.
  • Compact Size: Offers advanced reasoning within a 4.5B parameter model, making it efficient for deployment.

Good For

  • Mathematical Problem-Solving: Handling arithmetic, algebra, word problems, and complex quantitative tasks.
  • STEM Reasoning: Addressing scientific questions and structured analytical problems across various STEM fields.
  • Logical Deduction: Breaking down complex questions, performing multi-step reasoning, and verifying intermediate conclusions.
  • General Instruction Following: Executing structured tasks and responding to diverse queries with analytical depth.
  • Analytical Assistance: Aiding in comparing approaches, explaining solutions, and organizing technical information effectively.

Finetunes
3 models