XIAOMI MiMo · Ace Data Cloud
MiMo 2.6 APIBring conversation, reasoningand tool collaboration
into your application
Use Xiaomi MiMo through the familiar Chat Completions interface.
From Flash for everyday conversations to Pro for complex reasoning, choose the right model for different tasks.
- Pro / Flash
- Two models, choose by task
- Chat Completions
- Use the familiar conversation format
- SSE
- Progressively display generated content
- Token
- Billed by actual usage
CAPABILITIES
From a single Q&A to multi-step collaboration
Combine conversation, reasoning, and function tools to involve the model in real tasks within your application.
Multi-turn text conversations
Continue discussing the same topic, connect context through conversation history, and build Q&A, writing, and content-processing features.
messagesEnable reasoning by task
Enable reasoning for complex problems and disable it for simple tasks. Reasoning content and final answers are returned separately for easier display and handling in your application.
thinking.typeStreaming output
Receive content incrementally through SSE and update the interface during generation. When you need to track usage, you can request the final usage event.
stream: trueFunction tool calling
Declare retrieval, query, or business functions as tools. The model requests a call, your application executes it, and returns the result.
tools · tool_callsReuse existing SDKs
Continue using the OpenAI Chat Completions client; simply configure the MiMo service endpoint, API key, and model name to make requests.
client.chat.completions.createManage output and usage
Set a total output budget for reasoning and answers, view input, output, and cache-hit usage separately, and assess the cost of each task type.
max_completion_tokens · usageDEVELOPER EXPERIENCE
Familiar SDK,
new model choices
Start with a text request and integrate MiMo into existing applications. Add parameters for reasoning, streaming output, or function tools as needed for each task.
- Service endpoint
https://api.acedata.cloud/mimo- Model selection
mimo-v2.6-proormimo-v2.6-flash- Credential management
- Enable the corresponding service and store the API key in server-side environment variables.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ACEDATACLOUD_API_KEY"],
base_url="https://api.acedata.cloud/mimo",
)
response = client.chat.completions.create(
model="mimo-v2.6-pro",
messages=[
{"role": "user", "content": "解释彩虹形成的原因。"}
],
max_completion_tokens=256,
extra_body={"thinking": {"type": "disabled"}},
)
print(response.choices[0].message.content)
print(response.usage)This example disables reasoning and demonstrates a basic text request. When enabling reasoning, set a sufficient output budget for both reasoning and the final answer.
USE CASES
Bring MiMo into your workflow
Start with user input and turn model capabilities into clear application features.
Q&A and content assistant
Conduct multi-turn Q&A around product materials, help documentation, or user-provided context, assisting with drafting, rewriting, and organizing text.
Complex analysis and solution comparison
Enable reasoning for problems with multiple conditions and constraints to break down the problem, compare solutions, and then organize the final answer.
Programming and code understanding
Use code snippets and problem descriptions as text input to help explain logic, analyze errors, or draft modification suggestions for developers to verify.
Agents connected to business systems
Use custom functions to connect existing retrieval or query capabilities in your application. Return the results after executing tools so the model can continue processing the user's task.
MODEL LINEUP
Pro and Flash, choose by task
Use the same conversation format to configure models for different scenarios. Evaluate response quality, response time, and usage with your real tasks.
MiMo 2.6 Pro
Designed for complex analysis, reasoning, and multi-step tool tasks, suitable for problems that require deeper processing.
mimo-v2.6-pro- Multi-condition analysis and solution comparison
- Complex problem decomposition and reasoning
- Code understanding and multi-step tool collaboration
- Text conversations, streaming responses, and function tools
MiMo 2.6 Flash
Designed for everyday conversations, content processing, and lightweight tasks, suitable for common text interactions in applications.
mimo-v2.6-flash- Everyday Q&A and multi-turn conversations
- Text organization, summarization, and rewriting
- Lightweight analysis and function tool tasks
- Text conversations, streaming responses, and function tools
The two models are billed separately. When switching models, also verify the current pricing and your application's output budget.
GET STARTED
Integrate with your application in three steps
From obtaining credentials to processing results, follow a familiar development workflow.
Activate the service and obtain a key
Log in to the console, activate MiMo, and create the corresponding API credentials. Store the key on the server for your application to send requests.
Go to the application consoleConfigure the model and send requests
Set the service endpoint and Pro / Flash model identifier, then submit text through
View integration examplesmessages. Choose whether to enable reasoning or streaming responses based on the task.Process results and track usage
Read the response and final usage. When using tool calls, the application executes the function, returns the result, and retains the complete conversation history.
View API documentation
USAGE & PRICING
Clear pricing for every request based on actual usage
Input, cached input, and output are billed separately, providing a sound basis for model selection and cost estimation.
Standard input
Input tokens that do not hit the cache are billed at the input price of the selected model.
Cached input
Cache-hit input is billed at the cached price and is not charged again at the standard input price.
Response and reasoning output
Reasoning content is included in output tokens and is billed once together with the final response based on output usage.
Specific unit prices are subject to the current pricing page, and the actual cost of Credit depends on the purchased plan. Streaming requests are also billed based on final usage; when usage needs to be recorded, please enable stream_options.include_usage.
FAQ
Before integrating, you may want to know
Frequently asked questions about model selection, reasoning, tools, and billing. For complete parameters and error handling, please refer to the integration guide.
I already have an OpenAI SDK project. Can I integrate directly?
You can reuse the Chat Completions client. Set base_url to https://api.acedata.cloud/mimo, configure a key with MiMo service permissions, and use the exact model identifier. Refer to the MiMo API documentation for the specific supported parameters.
How should I choose between Pro and Flash?
Pro is designed for complex analysis, reasoning, and multi-step tool tasks; Flash is designed for everyday conversations, text processing, and lightweight tasks. We recommend comparing quality, response time, and token usage using the same real tasks before choosing the model that suits your application.
How do I enable reasoning? Is reasoning content billed twice?
Set thinking.type=enabled to enable reasoning and disabled to disable it. Reasoning and responses are returned separately in reasoning_content and content. Reasoning tokens are already included in output usage and will not be charged repeatedly.
Will the model directly execute my function tools?
No. The service currently supports tool declarations with type=function and tool_choice=auto. The model returns a call request, and your application executes the function and returns the result. When continuing, you must retain the reasoning content, complete tool calls, and corresponding call identifiers in the assistant message.
Are images, audio, web search, or other protocols supported?
The current service provides text Chat Completions, reasoning, SSE streaming output, and function tools. Image, audio, or video input, built-in web search, and Responses, Messages, Batch, and UltraSpeed are not yet available.
How should I handle rate limits or interrupted streams?
Use exponential backoff and retry when encountering 429; 503 may be returned when the service temporarily has no available capacity. If a streaming response has already started, first check the received content, usage, and call records before deciding whether to retry, to avoid repeated submissions.
EXPLORE MORE
Continue exploring other AI services
View the capabilities, integration methods, and current pricing of other model series.