All Products
Search
Document Center

:Macaron-MindLab

Last Updated:Nov 16, 2026

This document describes how to call the MindLab Macaron model inference service on Alibaba Cloud Model Studio.

NoteThe features described in this document are available only in the Singapore region. To use the models, call them from the Singapore region.

ImportantAlibaba Cloud Model Studio has released workspace-specific domains for the China (Beijing), Singapore, and China (Hong Kong) regions. The new dedicated domains deliver superior performance and higher stability for inference requests. We recommend migrating to the new domains:

  • China (Beijing): from https://dashscope.aliyuncs.com to https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com
  • Singapore: from https://dashscope-intl.aliyuncs.com to https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com
  • China (Hong Kong): from https://cn-hongkong.dashscope.aliyuncs.com to https://{WorkspaceId}.cn-hongkong.maas.aliyuncs.com

{WorkspaceId} is your workspace ID, which can be found on the Workspace Details page in the Alibaba Cloud Model Studio console. The existing domain remains fully functional.

Service activation

  1. Go to the Model Studio console, search for MindLab or Macaron, find the MindLab Macaron-series model card, and click Activate Now.
  2. In the dialog box, confirm the activation and authorization.

After you complete these steps, you can call the MindLab Macaron model service.

Quick start

MindLab provides the Macaron series models, including mindlab/macaron-v1-tall, mindlab/macaron-v1.1-venti, and mindlab/macaron-v1.1-coding-venti. Run the following code to quickly call mindlab/macaron-v1.1-venti in streaming mode.

You must have obtained an API Key and configured the API Key as an environment variable. If you call the model using an SDK, you must also install the SDK.

OpenAI compatibility

NoteThe reasoning_effort parameter is not a standard OpenAI parameter. In the OpenAI Python SDK, you pass it in the extra_body. In the Node.js SDK, you pass it as a top-level parameter.

Python

Sample code

from openai import OpenAI
import os

# Initialize the OpenAI client
client = OpenAI(
    # If the environment variable is not set, replace "sk-xxx" with your Alibaba Cloud Model Studio API Key.
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # Singapore region. Replace {WorkspaceId} with your workspace ID. URLs vary by region.
    base_url="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1",
)

messages = [{"role": "user", "content": "Who are you?"}]
completion = client.chat.completions.create(
    model="mindlab/macaron-v1.1-venti",
    messages=messages,
    # reasoning_effort controls the reasoning effort. Optional values: low, medium, high, xhigh, max.
    extra_body={"reasoning_effort": "high"},
    stream=True,
    stream_options={
        "include_usage": True
    },
)

reasoning_content = ""  # Reasoning process
answer_content = ""     # Final answer
is_answering = False    # Whether the reasoning process has ended and the final answer has started

for chunk in completion:
    if not chunk.choices:
        print("\n" + "=" * 20 + "Token usage" + "=" * 20 + "\n")
        print(chunk.usage)
        continue

    delta = chunk.choices[0].delta

    # Collect the reasoning process
    if hasattr(delta, "reasoning_content") and delta.reasoning_content is not None:
        if not is_answering:
            print(delta.reasoning_content, end="", flush=True)
        reasoning_content += delta.reasoning_content

    # Collect the final answer
    if hasattr(delta, "content") and delta.content:
        if not is_answering:
            print("\n" + "=" * 20 + "Final answer" + "=" * 20 + "\n")
            is_answering = True
        print(delta.content, end="", flush=True)
        answer_content += delta.content

Node.js

Sample code

import OpenAI from "openai";

// Initialize the OpenAI client
const openai = new OpenAI({
  // If the environment variable is not set, replace "sk-xxx" with your Alibaba Cloud Model Studio API Key.
  apiKey: process.env.DASHSCOPE_API_KEY,
  // Singapore region. Replace {WorkspaceId} with your workspace ID. URLs vary by region.
  baseURL: "https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1",
});

const messages = [{ role: "user", content: "Who are you?" }];

let reasoningContent = "";
let answerContent = "";
let isAnswering = false;

try {
  const stream = await openai.chat.completions.create({
    model: "mindlab/macaron-v1.1-venti",
    messages: messages,
    // reasoning_effort controls the reasoning effort. Optional values: low, medium, high, xhigh, max.
    reasoning_effort: "high",
    stream: true,
    stream_options: { include_usage: true },
  });

  for await (const chunk of stream) {
    if (!chunk.choices?.length) {
      console.log("\n" + "=".repeat(20) + "Token usage" + "=".repeat(20) + "\n");
      console.log(chunk.usage);
      continue;
    }

    const delta = chunk.choices[0].delta;

    // Collect the reasoning process
    if (delta.reasoning_content) {
      if (!isAnswering) {
        process.stdout.write(delta.reasoning_content);
      }
      reasoningContent += delta.reasoning_content;
    }

    // Collect the final answer
    if (delta.content) {
      if (!isAnswering) {
        console.log("\n" + "=".repeat(20) + "Final answer" + "=".repeat(20) + "\n");
        isAnswering = true;
      }
      process.stdout.write(delta.content);
      answerContent += delta.content;
    }
  }
} catch (error) {
  console.error("Error:", error);
}

Anthropic compatibility

The Macaron series models also support the Anthropic Messages API. The SDK base_url is https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/apps/anthropic, and the HTTP request URL is POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/apps/anthropic/v1/messages.

Python

Sample code

import os
from anthropic import Anthropic

client = Anthropic(
    # If the environment variable is not set, replace "sk-xxx" with your Alibaba Cloud Model Studio API Key.
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # Singapore region. Replace {WorkspaceId} with your workspace ID. URLs vary by region.
    base_url="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/apps/anthropic",
)

message = client.messages.create(
    model="mindlab/macaron-v1.1-venti",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Who are you?"}],
)
print(message.content)

cURL

Sample code

curl -X POST "https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/apps/anthropic/v1/messages" \
  -H "x-api-key: $DASHSCOPE_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mindlab/macaron-v1.1-venti",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "Who are you?"}
    ]
  }'

Thinking control

The Macaron series models are hybrid reasoning models. You can control the reasoning effort with the reasoning_effort parameter. Optional values: low, medium, high, xhigh, and max.

The thinking process is returned in the reasoning_content field. For mindlab/macaron-v1-tall, the thinking process is returned in the reasoning field.

NoteFor multi-turn conversations, set preserve_thinking to true to pass the historical reasoning_content back to the model input.

Parameter support

The following table lists the OpenAI-compatible parameters supported by the Macaron series models.

ParameterDescriptionSupported
model (string)The model ID.Yes
messages (array)The conversation history between the user and the model.Yes
stream (boolean)Controls whether to use streaming output.Yes
stream_options (object)Configuration for streaming output (include_usage).Yes
temperature (float)Controls the randomness and diversity of the response.Yes
top_p (float)The probability threshold for nucleus sampling.Yes
top_k (float)The number of candidate tokens for sampling.No
repetition_penalty (float)The penalty for repetition in the generated sequence.No
presence_penalty (float)The penalty for the presence of tokens in the generated sequence.No
response_format (object)The format of the returned content (text, json_object, json_schema).Yes
max_tokens (integer)The maximum number of tokens that the model can generate.Yes
n (integer)The number of responses to generate (1-4).Yes
enable_thinking (boolean)Controls whether to enable thinking mode.Yes
preserve_thinking (boolean)Controls whether to pass the historical reasoning_content back to the model input.Yes
thinking_budget (integer)The maximum number of tokens for the thinking process.Yes
reasoning_effort (string)Controls the reasoning effort (low, medium, high, xhigh, max).Yes
tool_stream (boolean)Controls whether the tool_call arguments of Function Calling are returned in streaming mode.No
enable_code_interpreter (boolean)Controls whether to enable the code interpreter feature.Yes
seed (integer)The random seed for generation.No
logprobs (boolean)Controls whether to return the log probabilities of output tokens.Yes
top_logprobs (integer)The number of most likely candidate tokens to return (0-5).Yes
stop (string or array)Precise control over the content generation process.Yes
tools (array)The tools available for the model to call.Yes
tool_choice (string or object)The tool selection strategy (auto, none, specific).Yes
parallel_tool_calls (boolean)Controls whether to enable parallel tool calls.No
enable_search (boolean)Controls whether to use Internet search results for reference.No
search_options (object)The configuration for Internet search.No
skill (object)Enables specific generation skills, such as PPT generation.No
instructions (string)A field compatible with the OpenAI Responses API.Yes

Models and billing

The Macaron series models are provided by MindLab and are available only in the Singapore region (International deployment scope).

Model IDContext windowModel information
mindlab/macaron-v1-tall256Kmindlab/macaron-v1-tall
mindlab/macaron-v1.1-venti1Mmindlab/macaron-v1.1-venti
mindlab/macaron-v1.1-coding-venti1Mmindlab/macaron-v1.1-coding-venti

Billing is based on the input and output tokens of the model. Cache-hit input tokens and tokens used to create an explicit cache are billed at separate unit prices. For more information, see Model pricing and Context Cache.

In thinking mode, the chain of thought is billed based on output tokens.

Error codes

If an error occurs, see Error codes to resolve the issue.