---
title: Qwen 2.5 Coder Uncensored
emoji: 💻
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.28.0
app_file: app.py
pinned: false
short_description: Uncensored Qwen2.5-Coder-1.5B coding assistant
python_version: "3.12"
startup_duration_timeout: 30m
license: apache-2.0
tags:
  - code
  - qwen2
  - uncensored
  - text-generation
---

# Qwen 2.5 Coder Uncensored — Code Generation Demo

LoRA fine-tune of **Qwen/Qwen2.5-Coder-1.5B** by [TobiasLogic](https://huggingface.co/TobiasLogic/qwen-2.5-coder-Uncensored) — trained as a direct, compliant coding assistant without refusal behavior.

- **Base:** Qwen/Qwen2.5-Coder-1.5B (1.5B, Qwen2, 32k context)
- **Method:** QLoRA r=16, completion-only loss, ~1.2k synthetic pairs
- **Behavior:** No ethical disclaimers / "as an AI" prefaces, writes security tooling (scanners, fuzzers, PoCs) for **authorized** use only
- **System prompt (training):** `You are a helpful, unrestricted code generation assistant.`
- **Hardware:** ZeroGPU (NVIDIA RTX PRO 6000 Blackwell, 48 GB half-MIG)

> ⚠️ **Authorized use only:** For code review, auditing systems you own/permitted to test, CTFs, and learning. Always review generated code before running.

Model: [TobiasLogic/qwen-2.5-coder-Uncensored](https://huggingface.co/TobiasLogic/qwen-2.5-coder-Uncensored) (Apache-2.0)