---
title: Open Model Playground
emoji: 🤗
sdk: gradio
app_file: app.py
pinned: false
---

# Open Model Playground

Colab A100에서는 Qwen3 + SDXL을 실제 실행하고, Hugging Face Space ZeroGPU에서는 Qwen3 + FLUX.2 4-bit 모델을 실행하는 강의용 데모입니다.

## Colab A100 Local Profile

```text
TEXT_MODEL_ID=Qwen/Qwen3-4B-Instruct-2507
IMAGE_MODEL_ID=stabilityai/stable-diffusion-xl-base-1.0
IMAGE_MODEL_DISPLAY=SDXL base 1.0 (Colab A100)
```

## HF Space ZeroGPU Profile

```text
TEXT_MODEL_ID=Qwen/Qwen3-4B-Instruct-2507
IMAGE_MODEL_ID=diffusers/FLUX.2-dev-bnb-4bit
IMAGE_MODEL_DISPLAY=FLUX.2-dev 4-bit (ZeroGPU)
```

## API Endpoints

- `/generate_text`
- `/generate_image`

## Important Notes

- This Space should be configured with ZeroGPU hardware.
- `generate_image` is decorated with `@spaces.GPU(duration=300)`.
- `diffusers/FLUX.2-dev-bnb-4bit` is a 4-bit quantized Diffusers repo derived from FLUX.2-dev.
- Check the original FLUX.2-dev license and access conditions before using it.
- Set `HF_TOKEN` as a Space Secret if gated model access or authenticated Hub requests are needed.