v1.1

Qwen-Image v1.1

Production-ready RunPod serverless endpoint for Qwen-Image (20B) text-to-image generation model.

Features

• Official Qwen-Image Model - 20B parameter MMDiT foundation model
• Exceptional Text Rendering - Superior quality for both English and Chinese text
• Auto-scaling - Scale to zero when idle (0-3 workers)
• GPU Optimized - Configured for A100 80GB and H100 series GPUs
• Apache 2.0 Licensed - Open source model

API Capabilities

• Text-to-image generation with customizable resolution
• Configurable inference steps (30-100)
• Seed-based reproducible generation
• Negative prompts support
• CFG scale control

Deployment Specs

• Min VRAM: 80GB recommended
• Container Disk: 5GB
• Network Volume: 100GB (persistent model cache) - ⚠️ REQUIRED
• Timeout: 600 seconds per job
• Supported GPUs: A100 80GB PCIe, H100 variants, RTX 6000 Blackwell series

RunPod Hub

This release is configured for RunPod Hub with:

• Production-ready tests for text rendering validation
• Pre-configured deployment presets (Default, Fast, High Quality)
• Complete API documentation

Example Usage

import runpod

endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")

request = {
    "input": {
        "prompt": "A serene mountain landscape with Chinese calligraphy",
        "width": 1024,
        "height": 1024,
        "num_inference_steps": 50
    }
}

result = endpoint.run_sync(request)

For detailed documentation, see the README.

v1.0.1

Qwen-Image v1.0.1

Production-ready RunPod serverless endpoint for Qwen-Image (20B) text-to-image generation model.

Features

• Official Qwen-Image Model - 20B parameter MMDiT foundation model
• Exceptional Text Rendering - Superior quality for both English and Chinese text
• Auto-scaling - Scale to zero when idle (0-3 workers)
• GPU Optimized - Configured for A100 80GB and H100 series GPUs
• Apache 2.0 Licensed - Open source model

API Capabilities

• Text-to-image generation with customizable resolution
• Configurable inference steps (30-100)
• Seed-based reproducible generation
• Negative prompts support
• CFG scale control

Deployment Specs

• Min VRAM: 80GB recommended
• Container Disk: 5GB
• Network Volume: 100GB (persistent model cache) - ⚠️ REQUIRED
• Timeout: 600 seconds per job
• Supported GPUs: A100 80GB PCIe, H100 variants, RTX 6000 Blackwell series

RunPod Hub

This release is configured for RunPod Hub with:

• Production-ready tests for text rendering validation
• Pre-configured deployment presets (Default, Fast, High Quality)
• Complete API documentation

Example Usage

import runpod

endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")

request = {
    "input": {
        "prompt": "A serene mountain landscape with Chinese calligraphy",
        "width": 1024,
        "height": 1024,
        "num_inference_steps": 50
    }
}

result = endpoint.run_sync(request)

For detailed documentation, see the README.

v1.0.0 - Initial Release

Qwen-Image Serverless v1.0.0

Production-ready RunPod serverless endpoint for Qwen-Image (20B) text-to-image generation model.

Features

• Official Qwen-Image Model - 20B parameter MMDiT foundation model
• Exceptional Text Rendering - Superior quality for both English and Chinese text
• Auto-scaling - Scale to zero when idle (0-3 workers)
• GPU Optimized - Configured for A100 80GB and H100 series GPUs
• Apache 2.0 Licensed - Open source model

API Capabilities

• Text-to-image generation with customizable resolution
• Configurable inference steps (30-100)
• Seed-based reproducible generation
• Negative prompts support
• CFG scale control

Deployment Specs

• Min VRAM: 80GB recommended
• Container Disk: 5GB
• Network Volume: 100GB (persistent model cache)
• Timeout: 600 seconds per job
• Supported GPUs: A100 80GB PCIe, H100 variants, RTX 6000 Blackwell series

RunPod Hub

This release is configured for RunPod Hub with:

• Production-ready tests for text rendering validation
• Pre-configured deployment presets (Default, Fast, High Quality)
• Complete API documentation

Example Usage

import runpod

endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")

request = {
    "input": {
        "prompt": "A serene mountain landscape with Chinese calligraphy",
        "width": 1024,
        "height": 1024,
        "num_inference_steps": 50
    }
}

result = endpoint.run_sync(request)

For detailed documentation, see the README.