How much does the Qwen-Image-2.1-Turbo API cost?

API pricing is $0.014133 per image in Beijing and other global regions, and $0.016 per image in Singapore and international regions.

What are the main features of Qwen-Image-2.1-Turbo?

It supports text-to-image generation, image-to-image editing, multi-reference inputs, local region control, and native transparency output.

Qwen-Image-2.1-Turbo Guide

Qwen-Image-2.1-Turbo API Pricing and Specs. A macro shot of a dense green GPU die reflecting amber status LEDs through a web of glowing fiber-optic cables inside a dark server chassis — GPUniq

Data as of October 15, 20264 sources

Drafted with AI tools and edited by a human. The figures and conclusions were checked by a GPUniq editor.

TL;DR

Qwen-Image-2.1-Turbo is a 7B visual generation model released by Alibaba on October 9, 2026. It uses an 8-step inference process and costs $0.014133 per image in Beijing, $0.016 in Singapore and international regions, and $0.014133 in other global regions.

What is Qwen-Image-2.1-Turbo?

Alibaba launched Qwen-Image-2.1-Turbo on October 9, 2026, as an accelerated version of the Qwen-Image-2.1 model. You can access it through the Alibaba Cloud Model Studio API or download the open-source weights from Hugging Face (Qwen/Qwen-Image-2.1-Turbo). It works with the QwenImage21Pipeline in standard open-source frameworks.

API Pricing

Region	Price, $/image
Beijing (China)	0.014133
Global (Singapore and international)	0.016000
Global (other regions)	0.014133

This structure positions it as the most cost-effective option in the Qwen-Image series.

Technical Specifications

Specification	Value
Model Architecture	7B visual generation
Inference Steps	8 denoising steps
Release Date	October 9, 2026
Predecessor Model	Qwen-Image-2.1

Capabilities

The model handles text-to-image generation and image-to-image editing, including multi-reference inputs and local region control like brushing or masking. It outputs native transparency and alpha channels based on your prompts while keeping the same 7B architecture as the base checkpoint.

Limitations

User reports indicate that detail and prompt adherence are somewhat lower than the full 40-step base model. The API and pipeline do not support the negative_prompt parameter, and you cannot customize the inference schedule beyond the saved 8 steps by default. It is published under the Qwen Research License, which may restrict certain commercial uses.

Methodology

Facts are compiled from the vendor's own announcement and documentation, linked below, at the time of writing; where the vendor hasn't published a number it is marked as such. GPUniq prices, when the model is already available here, come from our live catalog.