arsovskidev/Gemma-4-E4B-Claude-4.6-Opus-Reasoning-Distilled
arsovskidev/Gemma-4-E4B-Claude-4.6-Opus-Reasoning-Distilled is a 7.9 billion parameter language model, fine-tuned from Google's Gemma 4 E4B. It specializes in structured reasoning by distilling Chain-of-Thought samples from Claude 4.6 Opus, enabling it to plan responses within tags. This model excels at multi-step math, logic problems, and code decomposition, prioritizing deliberate thought over raw speed for complex analytical tasks.
Loading preview...
Overview
arsovskidev/Gemma-4-E4B-Claude-4.6-Opus-Reasoning-Distilled is a 7.9 billion parameter model, fine-tuned from google/gemma-4-E4B-it. Its core innovation lies in its training on approximately 2,300 high-quality Chain-of-Thought samples distilled from Claude 4.6 Opus. This process teaches the model to engage in structured, deliberate reasoning, often planning its approach within <think> tags before generating a final answer. The fine-tuning was performed using SFT + QLoRA (4-bit) with Unsloth, focusing on masking loss to responses only.
Key Capabilities
- Structured Reasoning: Learns to plan and break down problems step-by-step, reducing impulsive responses.
- Problem Solving: Enhanced ability in multi-step math and logic problems.
- Code Analysis: Proficient in code problem decomposition and debugging.
- Complex Prompt Analysis: Excels at structured analysis of intricate prompts.
Good For
- Tasks requiring explicit reasoning and problem-solving steps.
- Applications where the process of thought is as important as the final output.
- Analytical tasks in mathematics, logic, and programming.
Limitations
- Text-only: Does not support multimodal capabilities.
- Focused Scope: Trained on a relatively small dataset, making it a specialized reasoning fine-tune rather than a general-purpose upgrade.
- Hallucinations: Like all LLMs, it can still hallucinate, particularly on factual recall outside its training domain.