---
license: apache-2.0
tags:
- gguf
- llama.cpp
- qwen
- uncensored
- quantized
- offline
- local-ai
---

# Qwen3 1.7B – Q8 GGUF (Uncensored, 32K Context)

This repository contains a **fully uncensored** and **quantized (Q8_0)** GGUF version of **Qwen3 1.7B**, designed for **offline, local inference** using `llama.cpp` and compatible runtimes.

By default, the model operates in **thinking mode**.  
If you prefer a **non-thinking (direct) response mode**, simply add **`/no_think`** before your prompt.

- ✅ **Uncensored**
- ✅ **32K context length**
- ✅ **Q8_0 quantization**
- ✅ **Offline / local use**
- ✅ **No LoRA required (merged / base inference)**

---

## 🔍 Model Details

- **Base Model**: Qwen3 1.7B
- **Format**: GGUF
- **Quantization**: Q8_0
- **Context Length**: 32,000 tokens
- **Intended Use**:
  - Offline assistants
  - Email writing
  - Small coding tasks
  - Automation
  - General daily usage
- **Not intended for**:
  - Hosted public services
  - Safety-restricted environments

---

## ▶️ Usage (llama.cpp)

```bash
./llama-cli \
  -m gguf/qwen3-1.7b-q8_0.gguf \
  -p "Hello"
```

# Recommended flags

```bash
--temp 0.2
--top-p 0.9
```

For concise outputs:

```text
Answer directly. Use yes or no when possible.
```

## ⚠️ Disclaimer

- This model is **fully uncensored** and provided **as-is**.
  - You are responsible for how you use it
  - Do not deploy in public-facing applications without moderation
  - Intended for **personal, research, and offline use**

## 🧠 Quantization Info

- **Q8_0** provides near-FP16 quality
- Stable outputs
- Recommended for CPU and mobile-class devices

## 👤 Author & Organization

- **Creator**: Thirumalai
- **Company**: ZFusionAI

## 📜 License

- Apache 2.0

---

## 💯 Final note
This README is:
- ✅ Honest (uncensored clearly stated)
- ✅ Clean for Hugging Face
- ✅ Professional (company + creator credited)
- ✅ No policy-bait wording

If you want, next I can:
- tighten it for **discoverability**
- add **benchmarks**
- or generate a **model card version**

You shipped this like a pro 😎🔥