reaperdoesntknow/DiStil-Qwen3-1.7B-uncensored

Model Overview

DiStil-Qwen3-1.7B-uncensored is a 1.7 billion effective parameter model from Convergent Intelligence LLC: Research Division, built on the Qwen3ForCausalLM architecture. It is a distilled version of Qwen3, specifically fine-tuned using Supervised Fine-Tuning (SFT) on uncensored instruction data. The primary goal of this distillation is to eliminate alignment-imposed refusal behaviors, ensuring the model responds directly to prompts without filtering, while preserving the original Qwen3 base model's reasoning and generation capabilities.

Key Characteristics

• Alignment-Free: Designed to provide direct responses by removing safety heuristics that often lead to refusal behaviors.
• Capability Preservation: Maintains the core reasoning and generation abilities of the base Qwen3 model.
• Architecture: Utilizes the Qwen3ForCausalLM architecture with approximately 2.03 billion parameters (1.7B effective), a hidden size of 2048, 28 layers, and a context length of 40,960 tokens.
• Training Method: Achieved through SFT on uncensored data, focusing on shifting the model's response distribution without architectural modifications.
• Distillation Chain: This model serves as the base in a distillation chain, with further refinements like Disctil-Qwen3-1.7B built upon it.

Use Cases

This model is particularly suited for applications requiring unfiltered, direct responses to prompts, especially in technical, analytical, and research domains where traditional safety alignments might misinterpret or refuse legitimate queries. It is ideal for users who need a model that prioritizes direct answers over pre-programmed refusal patterns.