OpenAI-compatible Chat

Call models using the OpenAI-compatible Chat API, including input and output parameter descriptions and code examples.

Singapore

base_url for SDK: https://dashscope-intl.aliyuncs.com/compatible-mode/v1

HTTP endpoint: POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions

US (Virginia)

base_url for SDK: https://dashscope-us.aliyuncs.com/compatible-mode/v1

HTTP endpoint: POST https://dashscope-us.aliyuncs.com/compatible-mode/v1/chat/completions

China (Beijing)

base_url for SDK: https://dashscope.aliyuncs.com/compatible-mode/v1

HTTP endpoint: POST https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions

China (Hong Kong)

base_url for SDK: https://cn-hongkong.dashscope.aliyuncs.com/compatible-mode/v1

HTTP endpoint: POST https://cn-hongkong.dashscope.aliyuncs.com/compatible-mode/v1/chat/completions

Germany (Frankfurt)

base_url for SDK: ++https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/compatible-mode/v1++

HTTP endpoint: POST https://++{WorkspaceId}.eu-central-1.maas.aliyuncs.com++/compatible-mode/v1/chat/completions

Replace WorkspaceId with your actual Workspace ID. Prerequisites: Get an API key and set it as an environment variable. If you use the OpenAI SDK, install the SDK. Note

Regional API keys : API keys differ by region (Beijing vs Singapore/Virginia). See Create an API key for details. In code examples, replace DASHSCOPE_API_KEY with your actual key if not using environment variables.

## Request body

POST /chat/completions ** <svg> </svg> Debug </button>

## OpenAI-compatible API online debugging

<label> * International (Singapore) </label> <label> <input> Chinese Mainland (Beijing) </label>

POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions

<label> API Key Bearer Token </label> <input> <button> <svg> </svg></button> Get API Key for International (Singapore) region

<label>Request body</label> <button>Default</button> <button>Deep thinking</button> <button>Image input</button> <button>Tool calling</button> <button>Structured output</button> <button> <svg> <rect /> </svg> </button> <textarea>{ "model": "qwen-plus", "messages": [ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Who are you?" } ] }</textarea>

<button> Send request </button> <button> Clear response </button>

<label>Response result</label> <button>Raw response</button> <button>Parsed content</button> <button> <svg> <rect /> </svg> </button>

## Text input

## Python

python

import os
from openai import OpenAI

client = OpenAI(
 # If the environment variable is not set, replace the following line with: api_key="sk-xxx"
 api_key=os.getenv("DASHSCOPE_API_KEY"),
 # The following is the base_url for the Singapore region.
 base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
 model="qwen-plus", # Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 messages=\[
 {"role": "system", "content": "You are a helpful assistant."},
 {"role": "user", "content": "Who are you?"},
 \],
 # extra_body={"enable_thinking": False},
)
print(completion.model_dump_json())

## Java

java

// This code uses OpenAI SDK version 2.6.0
import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;

public class Main {
 public static void main(String\[\] args) {
 OpenAIClient client = OpenAIOkHttpClient.builder()
 .apiKey(System.getenv("DASHSCOPE_API_KEY"))
 // The following is the base_url for the Singapore region.
 .baseUrl("https://dashscope-intl.aliyuncs.com/compatible-mode/v1")
 .build();

ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
 .addUserMessage("Who are you?")
 .model("qwen-plus")
 .build();

try {
 ChatCompletion chatCompletion = client.chat().completions().create(params);
 System.out.println(chatCompletion);
 } catch (Exception e) {
 System.err.println("Error occurred: " + e.getMessage());
 e.printStackTrace();
 }
 }
}

## Node.js

nodejs

import OpenAI from "openai";

const openai = new OpenAI(
 {
 // If the environment variable is not set, replace the following line with: apiKey: "sk-xxx",
 apiKey: process.env.DASHSCOPE_API_KEY,
 // The following is the base_url for the Singapore region.
 baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
 }
);

async function main() {
 const completion = await openai.chat.completions.create({
 model: "qwen-plus", //Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 messages: \[
 { role: "system", content: "You are a helpful assistant." },
 { role: "user", content: "Who are you?" }
 \],
 });
 console.log(JSON.stringify(completion))
}

main();

## Go

package main

import (
"context"
"os"

"github.com/openai/openai-go"
"github.com/openai/openai-go/option"
)

func main() {
client := openai.NewClient(
 option.WithAPIKey(os.Getenv("DASHSCOPE_API_KEY")), // defaults to os.LookupEnv("OPENAI_API_KEY")
 // The following is the base_url for the Singapore region.
 option.WithBaseURL("https://dashscope-intl.aliyuncs.com/compatible-mode/v1/"),
)
chatCompletion, err := client.Chat.Completions.New(
 context.TODO(), openai.ChatCompletionNewParams{
 Messages: openai.F(
 \[\]openai.ChatCompletionMessageParamUnion{
 openai.UserMessage("Who are you?"),
 },
 ),
 Model: openai.F("qwen-plus"),
 },
)

if err != nil {
 panic(err.Error())
}

println(chatCompletion.Choices\[0\].Message.Content)
}

## C# (HTTP)

csharp

using System.Net.Http.Headers;
using System.Text;

class Program
{
 private static readonly HttpClient httpClient = new HttpClient();

static async Task Main(string\[\] args)
 {
 // If the environment variable is not set, replace the following line with: string? apiKey = "sk-xxx";
 string? apiKey = Environment.GetEnvironmentVariable("DASHSCOPE_API_KEY");

if (string.IsNullOrEmpty(apiKey))
 {
 Console.WriteLine("API Key is not set. Make sure the 'DASHSCOPE_API_KEY' environment variable is set.");
 return;
 }

// Set the request URL and content
 // The following is the base_url for the Singapore region.
 string url = "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions";
 // Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 string jsonContent = @"{
 ""model"": ""qwen-plus"",
 ""messages"": \[
 {
 ""role"": ""system"",
 ""content"": ""You are a helpful assistant.""
 },
 {
 ""role"": ""user"",
 ""content"": ""Who are you?""
 }
 \]
 }";

string result = await SendPostRequestAsync(url, jsonContent, apiKey);
 Console.WriteLine(result);
 }

private static async Task&lt;string&gt; SendPostRequestAsync(string url, string jsonContent, string apiKey)
 {
 using (var content = new StringContent(jsonContent, Encoding.UTF8, "application/json"))
 {
 httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", apiKey);
 httpClient.DefaultRequestHeaders.Accept.Add(new MediaTypeWithQualityHeaderValue("application/json"));

HttpResponseMessage response = await httpClient.PostAsync(url, content);
 if (response.IsSuccessStatusCode)
 {
 return await response.Content.ReadAsStringAsync();
 }
 else
 {
 return $"Request failed: {response.StatusCode}";
 }
 }
 }
}

## PHP (HTTP)

php

\<?php
// Set the request URL
// The following is the base_url for the Singapore region.
$url = 'https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions';
// If the environment variable is not set, replace the following line with: $apiKey = "sk-xxx";
$apiKey = getenv('DASHSCOPE_API_KEY');
$headers = \[
 'Authorization: Bearer '.$apiKey,
 'Content-Type: application/json'
\];
$data = \[
 // Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 "model" =\> "qwen-plus",
 "messages" =\> \[
 \[
 "role" =\> "system",
 "content" =\> "You are a helpful assistant."
 \],
 \[
 "role" =\> "user",
 "content" =\> "Who are you?"
 \]
 \]
\];
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, $url);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, json_encode($data));
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_HTTPHEADER, $headers);
$response = curl_exec($ch);
if (curl_errno($ch)) {
 echo 'Curl error: ' . curl_error($ch);
}
curl_close($ch);
echo $response;
?\>

## curl

API keys differ by region. For Beijing region, use https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions. See Get an API key.

curl

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \\
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \\
-H "Content-Type: application/json" \\
-d '{
 "model": "qwen-plus",
 "messages": \[
 {
 "role": "system",
 "content": "You are a helpful assistant."
 },
 {
 "role": "user",
 "content": "Who are you?"
 }
 \]
}'

## Streaming output

See Streaming output.

## Python

python

import os
from openai import OpenAI

client = OpenAI(
 # If the environment variable is not set, replace the following line with: api_key="sk-xxx"
 api_key=os.getenv("DASHSCOPE_API_KEY"),
 base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)
completion = client.chat.completions.create(
 model="qwen-plus", # Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 messages=\[{'role': 'system', 'content': 'You are a helpful assistant.'},
 {'role': 'user', 'content': 'Who are you?'},\],
 stream=True,
 stream_options={"include_usage": True}
 )
for chunk in completion:
 print(chunk.model_dump_json())

## Node.js

nodejs

import OpenAI from "openai";

const openai = new OpenAI(
 {
 apiKey: process.env.DASHSCOPE_API_KEY,
 baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
 }
);

async function main() {
 const completion = await openai.chat.completions.create({
 model: "qwen-plus", // Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 messages: \[
 {"role": "system", "content": "You are a helpful assistant."},
 {"role": "user", "content": "Who are you?"}
 \],
 stream: true,
 });
 for await (const chunk of completion) {
 console.log(JSON.stringify(chunk));
 }
}

main();

## curl

API keys differ by region. For Beijing region, use https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions. See Get an API key.

curl

curl --location "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions" \\
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \\
--header "Content-Type: application/json" \\
--data '{
 "model": "qwen-plus",
 "messages": \[
 {
 "role": "system",
 "content": "You are a helpful assistant."
 },
 {
 "role": "user",
 "content": "Who are you?"
 }
 \],
 "stream":true
}'

## Image input

For more image analysis options, see Image and video understanding.

## Python

python

import os
from openai import OpenAI

client = OpenAI(
 # If the environment variable is not set, replace the following line with: api_key="sk-xxx"
 api_key=os.getenv("DASHSCOPE_API_KEY"),
 base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)
completion = client.chat.completions.create(
 model="qwen-vl-plus", # Model list: https://www.alibabacloud.com/help/en/model-studio/models
 messages=\[{"role": "user","content": \[
 {"type": "image_url",
 "image_url": {"url": "https://dashscope.oss-cn-beijing.aliyuncs.com/images/dog_and_girl.jpeg"}},
 {"type": "text", "text": "What is this?"},
 \]}\]
 )
print(completion.model_dump_json())

## Node.js

nodejs

import OpenAI from "openai";

const openai = new OpenAI(
 {
 apiKey: process.env.DASHSCOPE_API_KEY,
 baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
 }
);

async function main() {
 const response = await openai.chat.completions.create({
 model: "qwen-vl-max", // Model list: https://www.alibabacloud.com/help/en/model-studio/models
 messages: \[{role: "user",content: \[
 { type: "image_url",image_url: {"url": "https://dashscope.oss-cn-beijing.aliyuncs.com/images/dog_and_girl.jpeg"}},
 { type: "text", text: "What is this?" },
 \]}\]
 });
 console.log(JSON.stringify(response));
}

main();

## curl

API keys differ by region. For Beijing region, use https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions. See Get an API key.

curl

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \\
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \\
-H 'Content-Type: application/json' \\
-d '{
 "model": "qwen-vl-plus",
 "messages": \[{
 "role": "user",
 "content": \[
 {"type": "image_url","image_url": {"url": "https://dashscope.oss-cn-beijing.aliyuncs.com/images/dog_and_girl.jpeg"}},
 {"type": "text","text": "What is this?"}
 \]}\]
}'

## Video input

Example: Using an image list as video input. For other methods (video files), see Visual understanding.

## Python

python

import os
from openai import OpenAI

client = OpenAI(
 # If the environment variable is not set, replace the following line with: api_key="sk-xxx"
 api_key=os.getenv("DASHSCOPE_API_KEY"),
 base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)
completion = client.chat.completions.create(
 # Model list: https://www.alibabacloud.com/help/en/model-studio/models
 model="qwen-vl-max",
 messages=\[{
 "role": "user",
 "content": \[
 {
 "type": "video",
 "video": \[
 "https://img.alicdn.com/imgextra/i3/O1CN01K3SgGo1eqmlUgeE9b_!!6000000003923-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i4/O1CN01BjZvwg1Y23CF5qIRB_!!6000000003000-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i4/O1CN01Ib0clU27vTgBdbVLQ_!!6000000007859-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i1/O1CN01aygPLW1s3EXCdSN4X_!!6000000005710-0-tps-3840-2160.jpg"\]
 },
 {
 "type": "text",
 "text": "Describe the process in this video."
 }\]}\]
)
print(completion.model_dump_json())

## Node.js

nodejs

// Make sure you have specified "type": "module" in package.json.
import OpenAI from "openai";

const openai = new OpenAI({
 // If the environment variable is not set, replace the following line with: apiKey: "sk-xxx",
 // API keys for the Singapore/Virginia and Beijing regions are different. To get an API key, see https://www.alibabacloud.com/help/en/model-studio/get-api-key
 apiKey: process.env.DASHSCOPE_API_KEY,
 baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
});

async function main() {
 const response = await openai.chat.completions.create({
 // Model list: https://www.alibabacloud.com/help/en/model-studio/models
 model: "qwen-vl-max",
 messages: \[{
 role: "user",
 content: \[
 {
 type: "video",
 video: \[
 "https://img.alicdn.com/imgextra/i3/O1CN01K3SgGo1eqmlUgeE9b_!!6000000003923-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i4/O1CN01BjZvwg1Y23CF5qIRB_!!6000000003000-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i4/O1CN01Ib0clU27vTgBdbVLQ_!!6000000007859-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i1/O1CN01aygPLW1s3EXCdSN4X_!!6000000005710-0-tps-3840-2160.jpg"
 \]
 },
 {
 type: "text",
 text: "Describe the process in this video."
 }
 \]}\]
 });
 console.log(JSON.stringify(response));
}

main();

## curl

The API keys for the Singapore/Virginia and Beijing regions are different. Create an API key. The following is the base_url for the Singapore region.

curl

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \\
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \\
-H 'Content-Type: application/json' \\
-d '{
 "model": "qwen-vl-max",
 "messages": \[
 {
 "role": "user",
 "content": \[
 {
 "type": "video",
 "video": \[
 "https://img.alicdn.com/imgextra/i3/O1CN01K3SgGo1eqmlUgeE9b_!!6000000003923-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i4/O1CN01BjZvwg1Y23CF5qIRB_!!6000000003000-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i4/O1CN01Ib0clU27vTgBdbVLQ_!!6000000007859-0-tps-3840-2160.jpg",
 "https://img.alicdn.com/imgextra/i1/O1CN01aygPLW1s3EXCdSN4X_!!6000000005710-0-tps-3840-2160.jpg"
 \]
 },
 {
 "type": "text",
 "text": "Describe the process in this video."
 }
 \]
 }
 \]
}'

## Tool calling

For the complete Function calling workflow code, see Function calling.

## Python

python

import os
from openai import OpenAI

client = OpenAI(
 # If the environment variable is not set, replace the following line with: api_key="sk-xxx"
 api_key=os.getenv("DASHSCOPE_API_KEY"),
 base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

tools = \[
 # Tool 1: Get current time (no parameters needed)
 {
 "type": "function",
 "function": {
 "name": "get_current_time",
 "description": "Useful when you want to know the current time.",
 "parameters": {}
 }
 },
 # Tool 2: Get weather (requires location parameter)
 {
 "type": "function",
 "function": {
 "name": "get_current_weather",
 "description": "Useful when you want to check the weather in a specific city.",
 "parameters": {
 "type": "object",
 "properties": {
 "location": {
 "type": "string",
 "description": "A city or district, such as Beijing, Hangzhou, or Yuhang District."
 }
 },
 "required": \["location"\]
 }
 }
 }
\]
messages = \[{"role": "user", "content": "What is the weather like in Hangzhou?"}\]
completion = client.chat.completions.create(
 model="qwen-plus", # Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 messages=messages,
 tools=tools
)

print(completion.model_dump_json())

## Node.js

nodejs

import OpenAI from "openai";

const openai = new OpenAI(
 {
 // If the environment variable is not set, replace the following line with: apiKey: "sk-xxx",
 apiKey: process.env.DASHSCOPE_API_KEY,
 baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
 }
);

const messages = \[{"role": "user", "content": "What is the weather like in Hangzhou?"}\];
const tools = \[
// Tool 1: Get current time (no parameters needed)
{
 "type": "function",
 "function": {
 "name": "get_current_time",
 "description": "Useful when you want to know the current time.",
 "parameters": {}
 }
},
// Tool 2: Get weather (requires location parameter)
{
 "type": "function",
 "function": {
 "name": "get_current_weather",
 "description": "Useful when you want to check the weather in a specific city.",
 "parameters": {
 "type": "object",
 "properties": {
 "location": {
 "type": "string",
 "description": "A city or district, such as Beijing, Hangzhou, or Yuhang District."
 }
 },
 "required": \["location"\]
 }
 }
}
\];

async function main() {
 const response = await openai.chat.completions.create({
 model: "qwen-plus", // Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 messages: messages,
 tools: tools,
 });
 console.log(JSON.stringify(response));
}

main();

## curl

The API keys for the Singapore/Virginia and Beijing regions are different. Create an API key. The following is the base_url for the Singapore region.

curl

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \\
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \\
-H "Content-Type: application/json" \\
-d '{
 "model": "qwen-plus",
 "messages": \[
 {
 "role": "system",
 "content": "You are a helpful assistant."
 },
 {
 "role": "user",
 "content": "What is the weather like in Hangzhou?"
 }
 \],
 "tools": \[
 {
 "type": "function",
 "function": {
 "name": "get_current_time",
 "description": "Useful when you want to know the current time.",
 "parameters": {}
 }
 },
 {
 "type": "function",
 "function": {
 "name": "get_current_weather",
 "description": "Useful when you want to check the weather in a specific city.",
 "parameters": {
 "type": "object",
 "properties": {
 "location":{
 "type": "string",
 "description": "A city or district, such as Beijing, Hangzhou, or Yuhang District."
 }
 },
 "required": \["location"\]
 }
 }
 }
 \]
}'

## Asynchronous invocation

python

import os
import asyncio
from openai import AsyncOpenAI
import platform

client = AsyncOpenAI(
 # If the environment variable is not set, replace the following line with: api_key="sk-xxx"
 # If you use a model in the China (Beijing) region, you need to use an API KEY for that region. Get it here: https://bailian.console.alibabacloud.com/?tab=model#/api-key
 api_key=os.getenv("DASHSCOPE_API_KEY"),
 base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

async def main():
 response = await client.chat.completions.create(
 messages=\[{"role": "user", "content": "Who are you?"}\],
 model="qwen-plus", # Model list: https://www.alibabacloud.com/help/en/model-studio/getting-started/models
 )
 print(response.model_dump_json())

if platform.system() == "Windows":
 asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
asyncio.run(main())

     &lt;b&gt;model** `&lt;i&gt;string*` **(Required)** The name of the model to use. Supported models include Qwen large language models (commercial and open source), Qwen-VL, Qwen-Coder, Qwen-Omni, and Qwen-Math, DeepSeek, Kimi, GLM, MiniMax. **For specific model names and billing information, check in the console.**     **messages** `*array*` **(Required)** The conversation history for the model, listed in chronological order.

Message types

System Message *object* (Optional) A system message that defines the role, tone, task objectives, or constraints for the large language model. Place it at the beginning of the messages array. ** Do not set a System Message for the QwQ model. A System Message has no effect on the QVQ model. <b>Properties**

content *string* (Required) A system instruction that defines the model's role, behavior, response style, and task constraints.

role *string* (Required) The role for a system message. The value is fixed as system.

User Message *object* (Required) A user message that passes questions, instructions, or context to the model. Properties

content *string or array* (Required) The message content. This is a string for text-only input. It is an array for multi-modal input, such as images, or if explicit caching is enabled. Properties for multi-modal models or when explicit caching is enabled

type *string* (Required) Valid values:

text Set to text for text input.
image_url Set to image_url for image input.
input_audio Set to input_audio for audio input.
video Set to video when the video input is an image list.
video_url Set to video_url for video file input. ** Only some Qwen-VL models accept video files as input. See Video understanding (Qwen-VL). The QVQ and Qwen-Omni models support direct video file input.

<b>text** *string* The input text. This parameter is required when type is text.

image_url *object* The input image information. This parameter is required when type is image_url. Properties

url *string* (Required) The URL or Base64 Data URL of the image. To pass a local file, see Image and video understanding.

input_audio *object* The input audio information. This parameter is required when type is input_audio. Properties

data *string* (Required) The URL or Base64 Data URL of the audio. To pass a local file, see Input Base64-encoded local file.

format *string* (Required) The format of the input audio, such as mp3 or wav.

video *array* The video information represented as an image list. This parameter is required when type is video. For usage instructions, see Video understanding (Qwen-VL), Video understanding (QVQ), or Video understanding (Qwen-Omni). Example value:

json

\[
 "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/xzsgiz/football1.jpg",
 "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/tdescd/football2.jpg",
 "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/zefdja/football3.jpg",
 "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/aedbqh/football4.jpg"
\]

video_url *object* The input video file information. This parameter is required when type is video_url. Qwen-VL can only understand the visual information in video files. Qwen-Omni can understand both the visual and audio information. Propertiesurl *string* (Required) The public network URL or Base64 Data URL of the video file. To input a local video file, see Input Base64-encoded local file.

fps *float* (Optional) The number of frames to extract per second. The value must be in the range of [0.1, 10]. The default value is 2.0. Function description Fps provides the following features:

For video file input, it controls the frame extraction frequency. One frame is extracted every <hetu> fps1 </hetu> seconds. ** Applicable to Qwen-VL and QVQ models.
It informs the model of the time interval between adjacent frames. This helps the model better understand the temporal dynamics of the video. This function applies to both video file and image list inputs. It is suitable for scenarios such as event time localization or segment content summarization. Supports the Qwen3.6, Qwen3.5, Qwen3-VL, Qwen2.5-VL, Qwen3.5-Omni, and QVQ models. A larger fps value is suitable for high-speed motion scenarios, such as sports events or action movies. A smaller fps value is suitable for long videos or content with static scenes.

<b>Example values**

For image list input: {"video":\["https://xx1.jpg",...,"https://xxn.jpg"\],"fps":2}
For video file input: {"video": "https://xx1.mp4","fps":2}

min_pixels *integer* (Optional) Sets the minimum pixel threshold for an input image or video frame. If an input image or video frame has a pixel count less than min_pixels, it is scaled up until the total pixel count exceeds min_pixels. Applicable to Qwen-VL and QVQ models. Value range

For image input: <li> Qwen3.6, Qwen3.5, and Qwen3-VL: The default and minimum value is 65536.
Qwen3.5-Omni: The default and minimum value is 24576.
qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: The default and minimum value is 4096.
Other qwen-vl-plus models, other qwen-vl-max models, open source Qwen2.5-VL series, and QVQ series models: The default and minimum value is 3136. </li>
For video file or image list input: <li> Qwen3.6, Qwen3.5-Omni, Qwen3.5, Qwen3-VL (including commercial and open source versions), qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: The default value is 65536. The minimum value is 4096.
Other qwen-vl-plus models, other qwen-vl-max models, open source Qwen2.5-VL series, and QVQ series models: The default value is 50176. The minimum value is 3136. </li>

Example values

For image input: {"type": "image_url","image_url": {"url":"https://xxxx.jpg"},"min_pixels": 65536}
For video file input: {"type": "video_url","video_url": {"url":"https://xxxx.mp4"},"min_pixels": 65536}
For image list input: {"type": "video","video": \["https://xx1.jpg",...,"https://xxn.jpg"\],"min_pixels": 65536}

max_pixels *integer* (Optional) Sets the maximum pixel threshold for an input image or video frame. If an input image or video has a pixel count within the \[min_pixels, max_pixels\] range, the model processes the original image. If the pixel count exceeds max_pixels, the image is scaled down until its total pixel count is less than max_pixels. Applicable to Qwen-VL and QVQ models. Value range

For image input: max_pixels is related to whether the vl_high_resolution_images parameter is enabled. <li> If vl_high_resolution_images is False: <li> Qwen3.6, Qwen3.5, and Qwen3-VL: The default value is 2621440. The maximum value is 16777216.
Qwen3.5-Omni: The default value is 1310720. The maximum value is 16777216.
qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: The default value is 1310720. The maximum value is 16777216.
Other qwen-vl-plus models, other qwen-vl-max models, open source Qwen2.5-VL series, and QVQ series models: The default value is 1003520. The maximum value is 12845056. </li>
If vl_high_resolution_images is True: <li> Qwen3.6, Qwen3.5-Omni, Qwen3.5, Qwen3-VL, qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: max_pixels is invalid. The maximum pixel count for an input image is fixed at 16777216.
Other qwen-vl-plus models, other qwen-vl-max models, open source Qwen2.5-VL series, and QVQ series models: max_pixels is invalid. The maximum pixel count for an input image is fixed at 12845056. </li> </li>
For video file or image list input: <li> Qwen3.6 series, Qwen3.5-Omni, Qwen3.5 series, close-source Qwen3-VL series, qwen3-vl-235b-a22b-thinking, and qwen3-vl-235b-a22b-instruct: The default value is 655360. The maximum value is 2048000.
Other Qwen3-VL open source models, qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: The default value is 655360. The maximum value is 786432.
Other qwen-vl-plus models, other qwen-vl-max models, open source Qwen2.5-VL series, and QVQ series models: The default value is 501760. The maximum value is 602112. </li>

Example values

For image input: {"type": "image_url","image_url": {"url":"https://xxxx.jpg"},"max_pixels": 8388608}
For video file input: {"type": "video_url","video_url": {"url":"https://xxxx.mp4"},"max_pixels": 655360}
For image list input: {"type": "video","video": \["https://xx1.jpg",...,"https://xxn.jpg"\],"max_pixels": 655360}

total_pixels *integer* (Optional) Limits the total number of pixels for all frames extracted from a video (pixels per frame × total number of frames). If the video's total pixel count exceeds this limit, the system scales down the video frames. However, it ensures that the pixel count of any single frame remains within the \[min_pixels, max_pixels\] range. Applicable to Qwen-VL and QVQ models. For long videos with many extracted frames, lower this value to reduce token consumption and processing time. However, this may cause a loss of image detail. Value range

Qwen3.6 series and Qwen3.5 series: The default and maximum value is 819200000. This value corresponds to 800000 image tokens (1 image token per 32×32 pixels).
Qwen3.5-Omni: The default and maximum value is 184549376. This value corresponds to 180224 image tokens (1 image token per 32×32 pixels).
Close-source Qwen3-VL series, qwen3-vl-235b-a22b-thinking, and qwen3-vl-235b-a22b-instruct: The default and maximum value is 134217728. This value corresponds to 131072 image tokens (1 image token per 32×32 pixels).
Other Qwen3-VL open source models, qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: The default and minimum value is 67108864. This value corresponds to 65536 image tokens (1 image token per 32×32 pixels).
Other qwen-vl-plus models, other qwen-vl-max models, open source Qwen2.5-VL series, and QVQ series models: The default and minimum value is 51380224. This value corresponds to 65536 image tokens (1 image token per 28×28 pixels).

Example values

For video file input: {"type": "video_url","video_url": {"url":"https://xxxx.mp4"},"total_pixels": 134217728}
For image list input: {"type": "video","video": \["https://xx1.jpg",...,"https://xxn.jpg"\],"total_pixels": 134217728}

cache_control *object* (Optional) Enables explicit caching. See Explicit cache. Propertiestype *string* (Required) Only ephemeral is supported.

role *string* (Required) The role for a user message. The value is fixed as user.

Assistant Message *object* (Optional) The model's reply. This is typically sent back to the model as context in a multi-turn conversation. Properties

content *string* (Optional) The text content of the model's reply. If tool_calls is included, content can be empty. Otherwise, content is required.

role *string* (Required) The role for an assistant message. The value is fixed as assistant.

partial *boolean* (Optional) Defaults to: false Specifies whether to enable partial mode. Valid values:

true
false See partial mode.

tool_calls *array* (Optional) The tool and input parameter information returned after a function call is initiated. It contains one or more objects and is retrieved from the tool_calls field of the previous model response. Properties

id *string* (Required) The ID of the tool response.

type *string* (Required) The tool type. Currently, only function is supported.

function *object* (Required) The tool and input parameter information. Properties

name *string* (Required) The tool name.

arguments *string* (Required) The input parameter information, formatted as a JSON string.

index *integer* (Required) The index of the current tool information in the tool_calls array.

Tool Message *object* (Optional) The output information of the tool. Properties

content *string* (Required) The output content of the tool function. It must be a string. If the tool returns structured data, such as JSON, serialize it into a string.

role *string* (Required) The value is fixed as tool.

tool_call_id *string* (Required) The ID returned after a function call is initiated. Retrieve it from completion.choices[0].message.tool_calls[$index].id. This ID marks the tool corresponding to the Tool Message.

 **stream** `*boolean*` (Optional. Defaults to `false`.) Enables streaming output mode. See Streaming output. Valid values:

false: The model returns the complete content at once after generation.
true: The model generates and sends output incrementally. A data block (chunk) is returned as soon as part of the content is generated. You can read these chunks in real time to assemble the full reply. You can set this parameter to true to improve the reading experience and reduce the risk of timeouts.

**Note ** For non-streaming calls, if the request does not complete within 300 seconds, the service will interrupt the request and return the content generated so far (instead of an error). For scenarios with long outputs, use streaming calls. For more information, see the timeout section in Text generation model overview.

 **stream_options** `*object*`* *(Optional) Configuration options for streaming output. This parameter is effective only when `stream` is set to `true`.

Properties

include_usage *boolean** *(Optional. Defaults to false) Specifies whether to include token consumption information in the last data block of the response. Valid values:

true
false ** For streaming output, token consumption information is included only in the last data block of the response.
<b>modalities *array *(Optional) Defaults to: \["text"\] Specifies the modalities of the output data. This parameter applies only to Qwen-Omni models. See Non-real-time (Qwen-Omni). Valid values:
\["text","audio"\]
\["text"\]audio *object* (Optional) The voice and format of the output audio. This parameter applies only to Qwen-Omni models, and you must set the modalities parameter to \["text","audio"\]. See Non-real-time (Qwen-Omni). Properties

voice *string* (Required) The voice used for the output audio. See Non-real-time (Qwen-Omni).

format *string* (Required) The format of the output audio. Only wav is supported.

 **temperature** `*float*` (Optional) The sampling temperature controls the diversity of the generated text. Higher values increase diversity, while lower values make the output more deterministic. The value must be greater than or equal to 0 and less than 2. Both the temperature and top_p parameters control the diversity of the generated text. Set only one of them. See Overview. ** Do not modify the default temperature value for QVQ models.      &lt;b&gt;top_p** `*float*` (Optional) The probability threshold for nucleus sampling. This parameter controls the diversity of the text that the model generates. A higher \`top_p\` value produces more diverse text. A lower \`top_p\` value produces more deterministic text. Value range: (0, 1.0\] Both \`temperature\` and \`top_p\` control the diversity of the generated text. Set only one of these parameters. See Overview. ** Do not modify the default \`top_p\` value for QVQ models.      &lt;b&gt;top_k** `*integer*` (Optional)  Specifies the number of candidate tokens to use for sampling during generation. A larger value produces more random output, whereas a smaller value produces more deterministic output. If set to `null` or a value greater than 100, the `top_k` strategy is disabled and only the `top_p` strategy takes effect. The value must be an integer greater than or equal to 0.

Default top_k values QVQ series, qwen-vl-plus-2025-07-10, and qwen-vl-plus-2025-08-15: 10. QwQ series: 40. Other qwen-vl-plus series, models before qwen-vl-max-2025-08-13, qwen2.5-omni-7b: 1. Qwen3-Omni-Flash series: 50. All other models: 20. GLM series (Alibaba Cloud direct): 20; DeepSeek/Kimi/MiniMax series do not support the top_k parameter.

** This parameter is not a standard OpenAI parameter. When using the Python SDK, include it in the <b>extra_body** object. Configure it as follows: extra_body={"top_k":xxx}. ** You must not change the default top_k value for QVQ models. <b>repetition_penalty *float* (Optional) Controls the repetitiveness of consecutive sequences during generation. A higher repetition_penalty reduces repetition, while 1.0 means no penalty. The value must be greater than 0, with no strict upper limit. ** This parameter is not a standard OpenAI parameter. When using the Python SDK, include it in the <b>extra_body object. Configure it as follows: extra_body={"repetition_penalty":xxx}. ** When using the qwen-vl-plus_2025-01-25 model for text extraction, we recommend setting repetition_penalty to 1.0. Do not modify the default repetition_penalty value for QVQ models. <b>presence_penalty** *float* (Optional) Controls how strongly the model avoids repeating content. Valid values: -2.0 to 2.0. Positive values reduce repetition. Negative values increase it. For scenarios that require diversity and creativity, such as creative writing or brainstorming, increase this value. For scenarios that require consistency and terminological accuracy, such as technical documents or formal text, decrease this value. Default presence_penalty values Qwen3.6 (non-thinking mode), Qwen3.5-Omni, Qwen3.5 (non-thinking mode), qwen3-max-preview (thinking mode), Qwen3 (non-thinking mode), Qwen3-Instruct series, qwen3-0.6b/1.7b/4b (thinking mode), QVQ series, qwen-max, qwen-max-latest, qwen-max-latest, qwen2.5-vl series, qwen-vl-max series, qwen-vl-plus, Qwen3-VL (non-thinking): 1.5. qwen-vl-plus-latest, qwen-vl-plus-2025-08-15: 1.2. qwen-vl-plus-2025-01-25: 1.0. qwen3-8b/14b/32b/30b-a3b/235b-a22b (thinking mode), qwen-plus/qwen-plus-latest/2025-04-28 (thinking mode), qwen-turbo/qwen-turbo/2025-04-28 (thinking mode): 0.5. All other models: 0.0. DeepSeek series (Alibaba Cloud direct): deepseek-r1, deepseek-r1-0528, deepseek-r1-distill-qwen distill series: 1; Kimi series (Alibaba Cloud direct): kimi-k2.6, kimi-k2.5: 0.0; Kimi series (Moonshot AI direct): 0.0; MiniMax series (Alibaba Cloud direct): MiniMax-M2.5, MiniMax-M2.1: 0.0; Other DeepSeek/Kimi/GLM/MiniMax models have no default value.

How it works When the parameter value is positive, the model penalizes tokens that already appear in the generated text. The penalty does not depend on how many times a token appears. This reduces the likelihood of those tokens reappearing, which decreases repetition and increases lexical diversity.

Example Prompt: Translate this sentence into English: "Esta película es buena. La trama es buena, la actuación es buena, la música es buena, y en general, toda la película es simplemente buena. Es realmente buena, de hecho. La trama es tan buena, y la actuación es tan buena, y la música es tan buena." Parameter value 2.0: This movie is very good. The plot is great, the acting is great, the music is also very good, and overall, the whole movie is incredibly good. In fact, it is truly excellent. The plot is very exciting, the acting is outstanding, and the music is so beautiful. Parameter value 0.0: This movie is good. The plot is good, the acting is good, the music is also good, and overall, the whole movie is very good. In fact, it is really great. The plot is very good, the acting is also very outstanding, and the music is also excellent. Parameter value -2.0: This movie is very good. The plot is very good, the acting is very good, the music is also very good, and overall, the whole movie is very good. In fact, it is really great. The plot is very good, the acting is also very good, and the music is also very good.

** When using the qwen-vl-plus-2025-01-25 model for text extraction, set presence_penalty to 1.5. Do not modify the default presence_penalty value for QVQ models. <b>response_format** *object* (Optional. Defaults to {"type": "text"}) The format of the response content. Valid values:

{"type": "text"}: Returns a plain text response.
{"type": "json_object"}: Returns a JSON string that conforms to standard JSON syntax. ** See Structured output. If you specify {"type": "json_object"}, explicitly instruct the model to output JSON in the prompt, such as by adding "Please output in JSON format." Otherwise, an error occurs. For supported models, see Structured output. <b>Properties**

type *string* (Required) The format of the response content. Valid values:

text: Returns a plain text response.
json_object: Returns a JSON string that conforms to standard JSON syntax.
max_tokens *integer* (Optional) The maximum number of tokens in the response. Generation stops when this limit is reached, and the finish_reason field in the response is set to length. The default and maximum values correspond to the model's maximum output length. Check in the console. You can use this parameter to control output length in scenarios such as generating summaries or keywords, or to reduce costs and shorten response time. When max_tokens is triggered, the finish_reason field in the response is set to length. ** max_tokens does not limit the length of the chain-of-thought.

<b>vl_high_resolution_images** *boolean* (Optional. Defaults to false.) Increases the maximum pixel limit for input images to the pixel value corresponding to 16384 tokens. See Processing high-resolution images.

vl_high_resolution_images: true: Uses a fixed-resolution strategy and ignores the max_pixels setting. If an image exceeds this resolution, its total pixel count is downscaled to meet the limit. Click to view the pixel limits for each model When vl_high_resolution_images is true, different models have different pixel limits: <li> Qwen3.6 series, Qwen3.5 series, Qwen3-VL series, qwen-vl-max, qwen-vl-max-latest, qwen-vl-max-0813, qwen-vl-plus, qwen-vl-plus-latest, qwen-vl-plus-0815: 16,777,216 (each Token corresponds to 32×32 pixels, i.e., 16,384×32×32)
QVQ series and other Qwen2.5-VL series models: 12,845,056 (each Token corresponds to 28×28 pixels, i.e., 16,384×28×28)

</li>

If vl_high_resolution_images is false, the actual pixel limit is determined by max_pixels. If an input image exceeds max_pixels, it is downscaled to fit within max_pixels. The default pixel limits for models match the default value of max_pixels.

** This parameter is not a standard OpenAI parameter. When making calls with the Python SDK, include this parameter in the <b>extra_body** object. The configuration is as follows: extra_body={"vl_high_resolution_images":xxx}. n *integer* (Optional. Defaults to 1) The number of responses to generate must be an integer in the range of 1-4. This is useful for scenarios that require multiple candidate responses, such as creative writing or ad copy. ** This parameter is supported only by the Qwen3 (non-thinking mode) models. If you pass the tools parameter, set n to 1. Increasing n increases output token consumption but does not affect input token consumption. <b>enable_thinking** *boolean* (Optional) Enables the thinking mode for hybrid thinking models. This mode is available for the Qwen3.6, Qwen3.5, Qwen3, Qwen3-Omni-Flash, and Qwen3-VL models, as well as the DeepSeek-V4-Pro/V4-Flash series, DeepSeek-V3.2/V3.2-exp/V3.1 series, Kimi-K2.6/K2.5 series, and GLM series. The DeepSeek-V4 series defaults to thinking mode. You can use the reasoning_effort parameter to adjust reasoning intensity. Valid values:

true ** When enabled, the thinking content is returned in the reasoning_content field.
false Default values for different models: Supported models This parameter is not a standard OpenAI parameter. When you make a call using the Python SDK, place it in the <b>extra_body** object. It is configured as follows: extra_body={"enable_thinking": xxx}. preserve_thinking *boolean* (Optional) Defaults to: false Controls whether to add the reasoning_content from past assistant messages to the model input. Use this when the model needs to refer to its past reasoning. This parameter currently supports qwen3.6-max-preview, qwen3.6-plus, qwen3.6-plus-2026-04-02, and kimi-k2.6 (deployed on Alibaba Cloud Model Studio).
If past messages lack reasoning_content, enabling this parameter does not cause an error.
When enabled, reasoning_content is included in the input tokens and is billed. ** This parameter is not a standard OpenAI parameter. When using the Python SDK, include it in the <b>extra_body** object. Configure this parameter as follows: extra_body={"preserve_thinking": True}. thinking_budget *integer* (Optional) The maximum number of tokens for the thinking process. This applies to Qwen3.6, Qwen3.5, Qwen3-VL, and the commercial and open source versions of Qwen3 models. See Limit reasoning length. The default value is the model's maximum chain-of-thought length. Check in the console. ** This parameter is not a standard OpenAI parameter. When you use the Python SDK, place this parameter in the <b>extra_body** object. Configure the parameter as follows: extra_body={"thinking_budget": xxx}. reasoning_effort *string* (Optional) Defaults to: high Controls the reasoning intensity for DeepSeek-V4 series models. Valid values:
high: High-intensity reasoning
max: Maximum-intensity reasoning low and medium map to high, and xhigh maps to max. Available for deepseek-v4-pro and deepseek-v4-flash. ** This parameter is not a standard OpenAI parameter. When using the Python SDK, include it in the <b>extra_body** object. Configure the parameter as follows: extra_body={"reasoning_effort": "high"}. tool_stream *boolean* (Optional) Defaults to: false When enabled, Function Calling tool_call arguments are returned in a streaming incremental manner rather than all at once. This parameter only takes effect when stream=true. Available for glm-5.1, glm-5, glm-4.7, and glm-4.6. ** This parameter is not a standard OpenAI parameter. When using the Python SDK, include it in the <b>extra_body** object. Configure the parameter as follows: extra_body={"tool_stream": true}. enable_code_interpreter *boolean* (Optional. Defaults to false.) Specifies whether to enable the code interpreter feature. See Code Interpreter. Valid values:
true
false ** This parameter is not a standard OpenAI parameter. When you make calls using the Python SDK, include this parameter in the <b>extra_body** object. The configuration is as follows: extra_body={"enable_code_interpreter": xxx}. seed *integer* (Optional) The random number seed. This parameter ensures that results are reproducible. If you use the same seed value in a call and the other parameters remain unchanged, the model returns the same result whenever possible. Valid values: \[0,231-1\]. logprobs *boolean* (Optional) Defaults to false Specifies whether to return the log probabilities of the output tokens. Valid values:
true
false ** Content generated during the thinking phase ( reasoning_content) does not include log probabilities. <b>Supported models**
Qwen-plus series snapshots (excluding the stable model)
Qwen-turbo series snapshots (excluding the stable model)
Qwen3-vl-plus models (including the stable model)
Qwen3-vl-flash models (including the stable model)
Qwen3 open source models
top_logprobs *integer* (Optional. Defaults to 0) You can specify the number of most likely candidate tokens to return at each generation step. Valid values: 0 to 5 This parameter applies only if logprobs is set to true. stop *string or array* (Optional) This parameter specifies stop words. If a string or token_id specified in stop appears in the text generated by the model, generation stops immediately. Pass sensitive words to control the model's output. ** If stop is an array, do not use a token_id or a string as elements simultaneously. For example, \["Hello",104307\] is not a valid value. <b>tools** *array** *(Optional) An array of one or more tool objects that the model can call in function calling. See Function calling. When tools is set and the model determines that a tool needs to be called, the response returns tool information in the tool_calls field. Properties

type *string** *(Required) Tool type. Currently supports only function.

function *object** *(Required)Properties

name *string** *(Required) The tool name. It must contain only letters, digits, underscores (_), and hyphens (-). The name can be up to 64 tokens long.

description *string** *(Required) A description of the tool. This helps the model determine when and how to call the tool.

parameters *object** *(Optional) Defaults to {} The tool's parameters must be described using a valid JSON Schema. For more information about JSON Schema, see this link. If the parameters parameter is empty, this indicates that the tool has no input parameters, such as a time query tool. ** To improve the accuracy of tool calling, we recommend passing parameters.

 &lt;b&gt;tool_choice** `*string or object*` (Optional. Defaults to `auto`.) The tool selection policy. Use this parameter to force a tool call for certain types of questions, such as always using a specific tool or disabling all tools. Valid values:

auto The model automatically selects a tool.
none To disable tool calling, set the tool_choice parameter to none.
{"type": "function", "function": {"name": "the_function_to_call"}} To force a call to a specific tool, set the tool_choice parameter to {"type": "function", "function": {"name": "the_function_to_call"}}, where the_function_to_call is the name of the specified tool function. ** Models in thinking mode do not support forcing a call to a specific tool. <b>parallel_tool_calls** *boolean* (Optional. Defaults to false.) Specifies whether to enable parallel tool calling. See Parallel tool calling. Valid values:
true
falseenable_search *boolean* (Optional) Defaults to: false Enables web search. See Web search. Valid values:
true ** If web search is not triggered after enabling this parameter, optimize the prompt or set the forced_search parameter in search_options to enable forced search.
false Enabling the web search feature may increase token consumption. This parameter is not a standard OpenAI parameter. When calling with the Python SDK, include it in the <b>extra_body** object. Configure it as follows: extra_body={"enable_search": True}. search_options *object* (Optional) The web search strategy. See Web search. Properties

forced_search *boolean* (Optional) Defaults to: false Forces web search. This parameter takes effect only when enable_search is true. Valid values:

true: Forcefully enables the feature.
false: Lets the model decide whether to perform a web search.

search_strategy *string* (Optional) Defaults to: turbo The search scale strategy. This parameter takes effect only when enable_search is true. Valid values:

turbo (default): Balances response speed and search effectiveness. Suitable for most scenarios.
max: Uses a more comprehensive search strategy and calls multiple search engines to retrieve more detailed results. Response time may be longer.
agent: Calls the web search tool and the model multiple times to retrieve and integrate information across multiple rounds. ** Applicable only to qwen3.5-plus, qwen3.5-plus-2026-02-15, qwen3.5-flash, qwen3.5-flash-2026-02-23, qwen3-max, qwen3-max-2026-01-23, qwen3-max-2025-09-23, qwen3.5-omni-plus, qwen3.5-omni-plus-2026-03-15, qwen3.5-omni-flash, and qwen3.5-omni-flash-2026-03-15.
agent_max: Adds web extraction support to the agent strategy. See Web extractor. Applicable only to the thinking mode of qwen3-max and qwen3-max-2026-01-23.

<b>enable_search_extension** *boolean* (Optional) Defaults to: false Enables domain-specific search. This parameter takes effect only when enable_search is true. Valid values:

true
false

** This parameter is not a standard OpenAI parameter. When you call it using the Python SDK, include it in the <b>extra_body** object.Configure it as follows: extra_body={"search_options": xxx}.

## Chat response object (non-streaming output)

json

{
 "choices": \[
 {
 "message": {
 "role": "assistant",
 "content": "I am a large-scale language model developed by Alibaba Cloud. My name is Qwen."
 },
 "finish_reason": "stop",
 "index": 0,
 "logprobs": null
 }
 \],
 "object": "chat.completion",
 "usage": {
 "prompt_tokens": 3019,
 "completion_tokens": 104,
 "total_tokens": 3123,
 "prompt_tokens_details": {
 "cached_tokens": 2048
 }
 },
 "created": 1735120033,
 "system_fingerprint": null,
 "model": "qwen-plus",
 "id": "chatcmpl-6ada9ed2-7f33-9de2-8bb0-78bd4035025a"
}

id *string* The unique identifier for this request.

choices *array* An array of generated content from the model. Properties

finish_reason *string* The reason the model stopped generating output. The following scenarios apply:

stop: The model stopped naturally or because the stop input parameter was triggered.
length: The generation stopped because the output reached the maximum length.
tool_calls: The model stopped to call a tool.

index *integer* The index of this object in the choices array.

logprobs *object* Log probability information for tokens in the model's output. Properties

content *array* An array of tokens and their corresponding log probabilities. Properties

token *string* The text of the current token.

bytes *array* A list of raw UTF-8 bytes for the current token. This list enables accurate reconstruction of output content, such as emojis or Chinese characters.

logprob *float* The log probability of the current token. A return value of null indicates an extremely low probability.

top_logprobs *array* The most likely candidate tokens for the current token position. The number of candidates matches the top_logprobs request parameter. Each element contains: Properties

token *string* The candidate token text.

bytes *array* A list of raw UTF-8 bytes for the current token. This list enables accurate reconstruction of output content, such as emojis or Chinese characters.

logprob *float* The log probability of this candidate token. A return value of null indicates an extremely low probability.

message *object* The message generated by the model. Properties

content *string* The content of the model's response.

reasoning_content *string* The content of the model's chain-of-thought reasoning.

refusal *string* This field is always null.

role *string* The role of the message. The value is always assistant.

audio *object* This field is always null.

function_call (to be deprecated) *object* This field is always null. For function calls, use the tool_calls parameter instead.

tool_calls *array* Information about tools and their input parameters that the model generates after initiating a function call. Properties

id *string* The unique identifier for this tool response.

type *string* The type of the tool. Currently, only function is supported.

function *object* Information about the tool. Properties

name *string* The name of the tool.

arguments *string* The input parameters, formatted as a JSON string. ** Model outputs are non-deterministic. The output parameters might not match the function signature. Validate the parameters before calling the function.

<b>index** *integer* The index of this tool in the tool_calls array.

created *integer* The Unix timestamp, in seconds, when the request was created.

model *string* The model used for this request.

object *string* The value is always chat.completion.

service_tier *string* This field is currently fixed as null.

system_fingerprint *string* This field is currently fixed as null.

usage *object* Token consumption details for this request. Properties

completion_tokens *integer* The number of tokens in the model's output.

prompt_tokens *integer* The number of tokens in the input.

total_tokens *integer* The total number of tokens consumed, equal to the sum of prompt_tokens and completion_tokens.

completion_tokens_details *object* A fine-grained breakdown of output tokens. Properties

audio_tokens *integer* This field is currently set to null.

reasoning_tokens *integer* This field is currently set to null.

text_tokens *integer* The number of text tokens in the output.

prompt_tokens_details *object* A fine-grained breakdown of input tokens. Properties

audio_tokens *integer* This field is currently set to null.

cached_tokens *integer* The number of tokens that hit the cache. See context cache.

text_tokens *integer* The number of text tokens in the input.

image_tokens *integer* The number of image tokens in the input.

video_tokens *integer* The number of tokens for the input video file or image list.

cache_creation *object* Information about the creation of an explicit cache. Properties

ephemeral_5m_input_tokens *integer* The number of tokens used to create the explicit cache.

cache_creation_input_tokens *integer* The number of tokens used to create the explicit cache.

cache_type *string* When you use an explicit cache, the value is ephemeral. Otherwise, this field does not exist.

## Chat response chunk object (streaming output)

json

{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"","function_call":null,"refusal":null,"role":"assistant","tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"I am a ","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"large-scale ","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"language model ","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"from Alibaba Cloud. My name ","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"is Qwen","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":".","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":null,"index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[{"delta":{"content":"","function_call":null,"refusal":null,"role":null,"tool_calls":null},"finish_reason":"stop","index":0,"logprobs":null}\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":null}
{"id":"chatcmpl-e30f5ae7-3063-93c4-90fe-beb5f900bd57","choices":\[\],"created":1735113344,"model":"qwen-plus","object":"chat.completion.chunk","service_tier":null,"system_fingerprint":null,"usage":{"completion_tokens":17,"prompt_tokens":22,"total_tokens":39,"completion_tokens_details":null,"prompt_tokens_details":{"audio_tokens":null,"cached_tokens":0}}}

id *string* The unique identifier for this request. All chunks in the same response share the same id.

choices *array* An array of generated content objects. This array contains one or more objects. If you set the include_usage parameter to true, the choices array is empty in the final chunk. Properties

delta *object* The incremental content object for this chunk. Properties

content *string* The incremental message content.

reasoning_content *string* The incremental chain-of-thought content.

function_call *object* This value defaults to null. See the tool_calls parameter.

audio *object* The response generated by the Non-real-time (Qwen-Omni) model. Properties

data *string* The incremental Base64-encoded audio data.

expires_at *integer* The UNIX timestamp when the request was created.

refusal *object* This parameter is always null.

role *string* The role of the incremental message object. This property appears only in the first chunk.

tool_calls *array* Information about tools and input parameters that the model generates after a function call. Properties

index *integer* The index of the current tool in the tool_calls array.

id *string* The unique ID for this tool response.

function *object* Information about the invoked tool. Properties

arguments *string* Incremental information about the input parameters. Concatenate the arguments from all chunks to obtain the complete input parameters. ** Because large model responses have a degree of randomness, the output parameter information may not conform to the function signature. Validate the parameters before you call the function.

<b>name** *string* The tool name. This property appears only in the first chunk.

type *string* The tool type. Currently, only function is supported.

finish_reason *string* The model stops generating for one of the following reasons:

stop: The stop input parameter was triggered or output stopped naturally.
Before generation is complete, it is null.
length: The maximum output length was reached.
tool_calls: The model stopped to make a tool call.

index *integer* The index of the current response in the choices array. If the input parameter n is greater than 1, use this parameter to reconstruct the complete content for each response.

logprobs *object* Probability information for the current object. Properties

content *array* An array of tokens with associated log probability information. Properties

token *string* The current token.

bytes *array* A list of the raw UTF-8 bytes of the current token. This is used to accurately restore the output content and is helpful when you handle emojis and Chinese characters.

logprob *float* The log probability of the current token. A return value of null indicates an extremely low probability.

top_logprobs *array* The most likely tokens at the current position and their log probabilities. The number of elements matches the top_logprobs input parameter. Properties

token *string* The current token.

bytes *array* A list of the raw UTF-8 bytes of the current token. This is used to accurately restore the output content and is helpful when you handle emojis and Chinese characters.

logprob *float* The log probability of the current token. A return value of null indicates an extremely low probability.

created *integer* The UNIX timestamp when this request was created. All chunks in the same response share the same created timestamp.

model *string* The model used for this request.

object *string* The value is always chat.completion.chunk.

service_tier *string* This parameter is currently fixed as null.

system_fingerprint *string* This parameter is currently fixed as null.

usage *object* Token usage statistics for this request. This object appears only in the final chunk when include_usage is true. Properties

completion_tokens *integer* The number of tokens in the model's output.

prompt_tokens *integer* The number of input tokens.

total_tokens *integer* The total number of tokens, equal to the sum of prompt_tokens and completion_tokens.

completion_tokens_details *object* Detailed breakdown of output tokens. Properties

audio_tokens *integer* The number of audio tokens in the output.

reasoning_tokens *integer* The number of tokens in the thinking process.

text_tokens *integer* The number of output text tokens.

prompt_tokens_details *object* A fine-grained breakdown of input tokens. Properties

audio_tokens *integer* The number of tokens in the input audio. ** The number of audio tokens in a video file is returned in this parameter.

<b>text_tokens** *integer* The number of tokens in the input text.

video_tokens *integer* The number of tokens for the input video, which can be an image list or a video file.

image_tokens *integer* The number of tokens in the input image.

cached_tokens *integer* The number of tokens that hit the cache. See Context cache.

cache_creation *object* Information about the creation of an explicit cache. Properties

ephemeral_5m_input_tokens *integer* The number of tokens used to create the explicit cache.

cache_creation_input_tokens *integer* The number of tokens used to create the explicit cache.

cache_type *string* The cache type. The value is fixed as ephemeral.

OpenAI-compatible Chat ​

Singapore ​

US (Virginia) ​

China (Beijing) ​

China (Hong Kong) ​

Germany (Frankfurt) ​

## Request body ​

## OpenAI-compatible API online debugging ​

## Text input ​

## Python ​

## Java ​

## Node.js ​

## Go ​

## C# (HTTP) ​

## PHP (HTTP) ​

## curl ​

## Streaming output ​

## Python ​

## Node.js ​

## curl ​

## Image input ​

## Python ​

## Node.js ​

## curl ​

## Video input ​

## Python ​

## Node.js ​

## curl ​

## Tool calling ​

## Python ​

## Node.js ​

## curl ​

## Asynchronous invocation ​

## Chat response object (non-streaming output) ​

## Chat response chunk object (streaming output) ​

OpenAI-compatible Chat

Singapore

US (Virginia)

China (Beijing)

China (Hong Kong)

Germany (Frankfurt)

## Request body

## OpenAI-compatible API online debugging

## Text input

## Python

## Java

## Node.js

## Go

## C# (HTTP)

## PHP (HTTP)

## curl

## Streaming output

## Python

## Node.js

## curl

## Image input

## Python

## Node.js

## curl

## Video input

## Python

## Node.js

## curl

## Tool calling

## Python

## Node.js

## curl

## Asynchronous invocation

## Chat response object (non-streaming output)

## Chat response chunk object (streaming output)