Question 1

What is Qwen Image 2?

Accepted Answer

Qwen Image 2.0 is a unified image generation and editing model built by Alibaba. It supports photorealistic outputs, accurate text rendering, and bilingual prompts (Chinese and English) in a single architecture.

Question 2

How much does Qwen Image 2 cost on Runflow?

Accepted Answer

Qwen Image 2 Standard is $0.035/image. The Pro variant (higher quality inference) is $0.075/image. Both are fixed per-image pricing — no GPU-second billing or cold start fees.

Question 3

Does Qwen Image 2 support Chinese prompts?

Accepted Answer

Yes. Qwen Image 2 natively understands both Chinese and English prompts. You can send prompts in either language without any preprocessing or translation step.

Question 4

What resolutions does Qwen Image 2 support?

Accepted Answer

Qwen Image 2 supports any resolution where total pixels are between 512×512 and 2048×2048. You can pass a preset (e.g. square_hd) or a custom width/height object via the image_size parameter.

Question 5

Can I use Qwen Image 2 for commercial projects?

Accepted Answer

Yes. Qwen Image 2 is available under a commercial license. You can use it in production SaaS products, client deliverables, and high-volume generation pipelines.

Question 6

How does Qwen Image 2 compare to FLUX and Midjourney?

Accepted Answer

Qwen Image 2 is competitive on photorealism and text rendering — areas where FLUX models have historically been weaker. At $0.035/image it's also cheaper than premium FLUX Pro variants. Detailed benchmark comparisons per use case are coming to Runflow.

Question 7

What's the difference between Qwen Image 2 and Qwen Image 2 Pro?

Accepted Answer

The Pro variant ($0.075/image) uses higher-quality inference for more detailed and nuanced outputs. For best Pro results, use 35+ inference steps. The standard variant ($0.035/image) is a strong default — use 25–28 steps for quality outputs, 15–20 for rapid drafts.

Question 8

How do I control quality vs speed with num_inference_steps?

Accepted Answer

Qwen Image 2 gives you precise control via num_inference_steps (default: 28). Use 15–20 steps for fast concept exploration, 25–28 for production-quality outputs, and 35+ steps when using the Pro endpoint for maximum fidelity.

Use Case	Assessment	Notes
Portrait	Strong	7B parameters deliver photorealistic portrait quality
Fashion	Moderate	Good detail, bilingual prompt support
Product	Strong	Accurate text rendering on product imagery
Creative	Moderate	Solid artistic flexibility with Chinese/English prompts

Parameter	Type	Required	Description
prompt	string	required	Text description of the image. Supports Chinese and English, up to 1,000 tokens.
image_size	object \| string	optional	Output size. Presets: "square", "square_hd", "landscape_4_3", "landscape_16_9", "portrait_4_3", "portrait_16_9". Or pass custom {width, height} — total pixels between 512×512 and 2048×2048. Default: "square_hd".
num_inference_steps	integer	optional	Denoising steps. Range: 1–50. Default: 28. Use 15–20 for drafts, 25–28 for quality, 35+ for final production.
guidance_scale	number	optional	How closely the model follows the prompt. Range: 0–20. Default: 5. Lower = more creative; higher = more literal.
num_images	integer	optional	Number of images to generate. Range: 1–4. Default: 1.
output_format	string	optional	Output format. Options: "jpeg", "png", "webp". Default: "png".
negative_prompt	string	optional	Content to avoid in the output. Max 500 characters.
enable_prompt_expansion	boolean	optional	Enable LLM prompt optimisation for better results. Default: true.
enable_safety_checker	boolean	optional	Enable content moderation for input and output. Default: true.
seed	integer	optional	Random seed for reproducibility (0–2147483647).
sync_mode	boolean	optional	Return media as data URI without storing in request history. Default: false.

Qwen Image 2 API

Qwen Image 2 model family by Alibaba

Qwen Image 2 (Text-to-Image)

Qwen Image 2 Pro (Text-to-Image)

Qwen Image 2 (Edit)

Qwen Image 2 Pro (Edit)

Benchmarked for your use case

API Parameters

Technical Specifications

Why run Qwen Image 2 on the Runflow API

Professional typography & text rendering

Bilingual prompt support

Tunable quality vs speed

Sentinel quality scoring

Explore other models

FLUX.1 [dev]

Seedream 4.5

Seedream 4

Nano Banana 2

GPT-Image 1.5

Nano Banana Pro

Frequently asked questions