Ayodele01/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled

Hugging Face
VISIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Mar 8, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

Ayodele01/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled is a 27 billion parameter Qwen 3.5-based language model fine-tuned for enhanced reasoning capabilities. Developed by Ayodele01, it leverages a cleaned subset of the Opus-4.6-Reasoning-3000x-filtered dataset, focusing on multi-step intermediate reasoning for mixed math and code-oriented prompts. This model is optimized for research and experimentation in reasoning-style text generation, providing both 16-bit merged weights and GGUF exports for flexible deployment.

Loading preview...

Model Overview

Ayodele01/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled is a 27 billion parameter language model built upon the unsloth/Qwen3.5-27B base. It has been specifically fine-tuned using a cleaned subset of the nohurry/Opus-4.6-Reasoning-3000x-filtered dataset, which comprises 2,183 conversations, primarily focusing on math (2,052 examples) and code (131 examples).

Key Capabilities

  • Enhanced Reasoning: The model is fine-tuned for long-form reasoning and answer generation, particularly for problems requiring multi-step intermediate thought processes.
  • Math and Code Proficiency: Its training data emphasizes mixed math and code-oriented prompts, making it suitable for tasks in these domains.
  • Explicit Reasoning Traces: The model is designed to emit visible <think> sections or long intermediate reasoning, reflecting its training on explicit reasoning traces.
  • Flexible Deployment: Available as 16-bit merged weights for Transformers/vLLM and GGUF exports (including q4_k_m, q8_0, q5_k_m quantizations) for llama.cpp-compatible runtimes.

Intended Use Cases

This model is ideal for:

  • Research and Experimentation: Exploring reasoning-style text generation.
  • Problem Solving: Tackling mixed math and code problems that benefit from detailed, step-by-step reasoning.
  • Educational Tools: Developing applications that require models to show their thought process.

Limitations: Quantized GGUF variants may differ from the 16-bit checkpoint, and no formal evaluation benchmarks are included in this release.

Finetunes
1 models