Qwen-Image-3.0 API

Nano Banana 2.1

Explore the Nano Banana 2.1 API.

xAI

Grok-Imagine-Image-2.0

Grok Imagine Image 2.0, its upgraded AI image generation and editing model

OpenAI

GPT Image 2

GPT Image 2 is openai state-of-the-art image generation model for fast, high-quality image generation and editing. It supports flexible image sizes and high-fidelity image inputs.

Gen4-image

Explore the Gen4-image API.

Gen4-image-turbo

Explore the Gen4-image-turbo API.

Nano Banana 2 lite

Gemini 3.1 Flash Lite Image model is an efficiency expert in the image generation family, designed for ultra-low latency and cost-effective image generation and modification.

Uptime

Review live heartbeat data, endpoint availability and observed response times before moving into production.

Updates

Follow meaningful availability, pricing and capability changes for Qwen-Image-3.0 without losing the stable model page.

Model changelog

One page that gets stronger over time.

Release notes, pricing changes, benchmark updates and migration guidance accumulate here while the canonical URL stays unchanged.

Pre-launch

Pre-launch model page published

A permanent model entity page is reserved for Qwen-Image-3.0; unconfirmed data remains clearly marked until release.

Live data

Pricing and limits synced from the catalog

Price, context, availability and limits are treated as dynamic properties instead of being embedded in the page title or model identity.

Aug 12, 2026

Permanent model entity created

The provider and model slug form a durable canonical URL; future content and data updates remain on this page.

Keep exploring

Compare models and continue your research

Coming soon

Aliyun

qwen-image-3.0

Qwen-Image-3.0 API

coming soon

Starting price

View pricing

Usage-based billing

API health

Available

Production endpoint

Response

Async ready

Observed API response

Released

Aug 11, 2026

Aliyun

Model brief

A stable model identity for search and evaluation, paired with live catalog data that can change without rewriting the page's core SEO structure.

Stable identity

What is Qwen-Image-3.0?

coming soon

Canonical model identity stays stable as pricing and limits change.

Live model data

Current catalog snapshot

Availability

Coming soon

Starting price

View pricing

Context

See capabilities

Catalog date

Aug 11, 2026

Overview

A practical overview of Qwen-Image-3.0, including important behavior and integration considerations.

What Is Qwen-Image-3.0?

What are the key features of Qwen-Image-3.0?

Multiple outputs: The API supports up to 6 output images per request, allowing users to generate several variations from the same prompt.

Qwen-Image-3.0 Technical Specifications

Specification	Qwen-Image-3.0
Developer	Alibaba Qwen Team
Model type	Image generation and image editing
Primary positioning	Balanced quality, speed, and cost
Model ID	qwen-image-3.0
Input modalities	Text and images
Output modality	Image
Generation modes	Text-to-image; image-to-image; instructed editing
Maximum prompt	Approximately 4.5K tokens
Reference images	1-3
Output pixel range	Total pixels from 512 x 512 to 2048 x 2048
Output format	PNG
Claimed small-text rendering	Approximately 10 px
Native visual-text languages	12
Standard rate limit	20 RPM in documented regions
Function calling	Not supported
Batch inference	Not supported
Fine-tuning	Not supported
CometAPI endpoints	/v1/images/generations; /v1/images/edits

Qwen-Image-3.0 Benchmark Performance

Model	T2I Elo	T2I Rank	Edit Elo	Edit Rank	API Price / 1K
GPT Image 2 (high)	1,367	#1	1,257	#4	$211
Nano Banana 2	1,319	#3	1,250	#7	$67
Qwen-Image-3.0-Pro	1,284	#9	1,250	#5	$43
Seedream 5.0 Pro	1,279	#10	1,247	#8	$90
Qwen-Image-3.0	1,275	#11	1,218	#15	$30

What the Results Mean

Standard versus Pro: the text-to-image gap is only 9 Elo, but Pro leads by 32 Elo in editing. The strongest reason to pay for Pro may therefore be edit quality rather than first-pass generation.
Against Seedream 5.0 Pro: Standard is only 4 Elo behind for text-to-image, while Pro is 5 Elo ahead. These small differences sit within the published uncertainty ranges and should be treated as a competitive cluster, not a definitive ordering.
Against Nano Banana 2: Standard trails by 44 Elo in text-to-image and 32 Elo in editing. Pro narrows the editing gap to a tie at 1,250 Elo.
Against GPT Image 2: GPT leads Standard by 92 Elo in text-to-image and 39 Elo in editing. Qwen's case is therefore price-performance and layout specialization, not universal quality leadership.
Compared with the earlier Qwen Image 2.0 Pro, the Standard 3.0 model gains 39 text-to-image Elo in the same leaderboard while the representative price falls from $75 to $30 per 1,000 images.

Qwen-Image-3.0 vs GPT Image 2 vs Nano Banana 2

Dimension	Qwen 3 Standard	Qwen 3 Pro	GPT Image 2	Nano Banana 2	Seedream 5 Pro
Positioning	Quality / speed / cost balance	Highest Qwen fidelity	Premium general image model	Fast, high-volume Gemini image model	Professional design and editing
T2I / Edit Elo	1,275 / 1,218	1,284 / 1,250	1,367 / 1,257	1,319 / 1,250	1,279 / 1,247
Cited output	About 2K	About 2K	Flexible sizes	Up to 4K	About 2K on CometAPI
Reference inputs	1-3	1-3	Supported	Multi-reference	Up to 10 on CometAPI
Typography angle	4.5K brief; 10px claim; 12 languages	Same core focus with higher fidelity	Strong multilingual text	Text plus search grounding	Dense infographics and spatial controls
Web grounding	No	No	Not a defining API feature	Google Search and Image Search	Access-path dependent
Best fit	Dense layouts at controlled cost	High-quality Qwen editing	Maximum general preference	Fast 4K and current-information workflows	Precise edits and multi-reference design

FAQ

API

Copy a working endpoint and code example, then open the complete API reference when you need every parameter.

One API, production-ready model access

Authenticate once, call the model endpoint and keep the same billing and observability workflow across providers.

Sample code and API for Qwen-Image-3.0

set -euo pipefail
: "${QWEN_IMAGE_INPUT_URL:?Set QWEN_IMAGE_INPUT_URL before running this example}"

generation_payload=$(jq -nc \
  --arg model "qwen-image-3.0" \
  '{
    model: $model,
    input: {
      messages: [{
        role: "user",
        content: [{
          text: "Create a cinematic watercolor of a moonlit harbor with glowing paper lanterns."
        }]
      }]
    },
    parameters: {
      negative_prompt: "blurry, low quality",
      prompt_extend: true,
      watermark: false,
      size: "1328*1328"
    }
  }')

echo "Generation response:"
curl --fail-with-body --silent --show-error --request POST \
  https://api.cometapi.com/v1/images/generations \
  --header "Authorization: Bearer $COMETAPI_KEY" \
  --header "Content-Type: application/json" \
  --data "${generation_payload}"
echo

edit_payload=$(jq -nc \
  --arg model "qwen-image-3.0" \
  --arg image "${QWEN_IMAGE_INPUT_URL}" \
  '{
    model: $model,
    input: {
      messages: [{
        role: "user",
        content: [
          {image: $image},
          {text: "Turn the scene into a moonlit watercolor while preserving the main composition."}
        ]
      }]
    },
    parameters: {
      n: 1,
      negative_prompt: "blurry, low quality",
      prompt_extend: true,
      watermark: false
    }
  }')

echo "Edit response:"
curl --fail-with-body --silent --show-error --request POST \
  https://api.cometapi.com/v1/images/edits \
  --header "Authorization: Bearer $COMETAPI_KEY" \
  --header "Content-Type: application/json" \
  --data "${edit_payload}"
echo

cURL Code Example

set -euo pipefail
: "${QWEN_IMAGE_INPUT_URL:?Set QWEN_IMAGE_INPUT_URL before running this example}"

generation_payload=$(jq -nc \
  --arg model "qwen-image-3.0" \
  '{
    model: $model,
    input: {
      messages: [{
        role: "user",
        content: [{
          text: "Create a cinematic watercolor of a moonlit harbor with glowing paper lanterns."
        }]
      }]
    },
    parameters: {
      negative_prompt: "blurry, low quality",
      prompt_extend: true,
      watermark: false,
      size: "1328*1328"
    }
  }')

echo "Generation response:"
curl --fail-with-body --silent --show-error --request POST \
  https://api.cometapi.com/v1/images/generations \
  --header "Authorization: Bearer $COMETAPI_KEY" \
  --header "Content-Type: application/json" \
  --data "${generation_payload}"
echo

edit_payload=$(jq -nc \
  --arg model "qwen-image-3.0" \
  --arg image "${QWEN_IMAGE_INPUT_URL}" \
  '{
    model: $model,
    input: {
      messages: [{
        role: "user",
        content: [
          {image: $image},
          {text: "Turn the scene into a moonlit watercolor while preserving the main composition."}
        ]
      }]
    },
    parameters: {
      n: 1,
      negative_prompt: "blurry, low quality",
      prompt_extend: true,
      watermark: false
    }
  }')

echo "Edit response:"
curl --fail-with-body --silent --show-error --request POST \
  https://api.cometapi.com/v1/images/edits \
  --header "Authorization: Bearer $COMETAPI_KEY" \
  --header "Content-Type: application/json" \
  --data "${edit_payload}"
echo

Python Code Example

import json
import os

import requests


COMETAPI_KEY = os.environ["COMETAPI_KEY"]
QWEN_IMAGE_INPUT_URL = os.environ["QWEN_IMAGE_INPUT_URL"]

headers = {
    "Authorization": f"Bearer {COMETAPI_KEY}",
    "Content-Type": "application/json",
}

generation_response = requests.post(
    "https://api.cometapi.com/v1/images/generations",
    headers=headers,
    json={
        "model": "qwen-image-3.0",
        "input": {
            "messages": [
                {
                    "role": "user",
                    "content": [
                        {
                            "text": "Create a cinematic watercolor of a moonlit harbor with glowing paper lanterns."
                        }
                    ],
                }
            ]
        },
        "parameters": {
            "negative_prompt": "blurry, low quality",
            "prompt_extend": True,
            "watermark": False,
            "size": "1328*1328",
        },
    },
    timeout=120,
)
generation_response.raise_for_status()
print("Generation response:")
print(json.dumps(generation_response.json(), ensure_ascii=False, indent=2))

edit_response = requests.post(
    "https://api.cometapi.com/v1/images/edits",
    headers=headers,
    json={
        "model": "qwen-image-3.0",
        "input": {
            "messages": [
                {
                    "role": "user",
                    "content": [
                        {"image": QWEN_IMAGE_INPUT_URL},
                        {
                            "text": "Turn the scene into a moonlit watercolor while preserving the main composition."
                        },
                    ],
                }
            ]
        },
        "parameters": {
            "n": 1,
            "negative_prompt": "blurry, low quality",
            "prompt_extend": True,
            "watermark": False,
        },
    },
    timeout=120,
)
edit_response.raise_for_status()
print("Edit response:")
print(json.dumps(edit_response.json(), ensure_ascii=False, indent=2))

JavaScript Code Example

const apiKey = process.env.COMETAPI_KEY;
const inputImageUrl = process.env.QWEN_IMAGE_INPUT_URL;

if (!apiKey || !inputImageUrl) {
  throw new Error("Set COMETAPI_KEY and QWEN_IMAGE_INPUT_URL before running this example.");
}

const headers = {
  Authorization: `Bearer ${apiKey}`,
  "Content-Type": "application/json",
};

const generationResponse = await fetch(
  "https://api.cometapi.com/v1/images/generations",
  {
    method: "POST",
    headers,
    body: JSON.stringify({
      model: "qwen-image-3.0",
      input: {
        messages: [
          {
            role: "user",
            content: [
              {
                text: "Create a cinematic watercolor of a moonlit harbor with glowing paper lanterns.",
              },
            ],
          },
        ],
      },
      parameters: {
        negative_prompt: "blurry, low quality",
        prompt_extend: true,
        watermark: false,
        size: "1328*1328",
      },
    }),
  },
);

if (!generationResponse.ok) {
  throw new Error(`Generation failed (${generationResponse.status}): ${await generationResponse.text()}`);
}

console.log("Generation response:");
console.log(JSON.stringify(await generationResponse.json(), null, 2));

const editResponse = await fetch(
  "https://api.cometapi.com/v1/images/edits",
  {
    method: "POST",
    headers,
    body: JSON.stringify({
      model: "qwen-image-3.0",
      input: {
        messages: [
          {
            role: "user",
            content: [
              { image: inputImageUrl },
              {
                text: "Turn the scene into a moonlit watercolor while preserving the main composition.",
              },
            ],
          },
        ],
      },
      parameters: {
        n: 1,
        negative_prompt: "blurry, low quality",
        prompt_extend: true,
        watermark: false,
      },
    }),
  },
);

if (!editResponse.ok) {
  throw new Error(`Edit failed (${editResponse.status}): ${await editResponse.text()}`);
}

console.log("Edit response:");
console.log(JSON.stringify(await editResponse.json(), null, 2));

Use cases

Use Qwen-Image-3.0 for production workflows that match its image capabilities, then compare alternatives before committing to a long-term integration.

Product visuals and campaign creative

Concept art and rapid visual exploration

Image editing, variation and asset production

Other models

Compare other models available through CometAPI for different quality, latency, capability and pricing trade-offs.

Open model comparison →

Nano Banana 2.1

Explore the Nano Banana 2.1 API.

xAI

Grok-Imagine-Image-2.0

Grok Imagine Image 2.0, its upgraded AI image generation and editing model

OpenAI

GPT Image 2

GPT Image 2 is openai state-of-the-art image generation model for fast, high-quality image generation and editing. It supports flexible image sizes and high-fidelity image inputs.

Gen4-image

Explore the Gen4-image API.

Gen4-image-turbo

Explore the Gen4-image-turbo API.

Nano Banana 2 lite

Gemini 3.1 Flash Lite Image model is an efficiency expert in the image generation family, designed for ultra-low latency and cost-effective image generation and modification.