Xiaomi MiMo V2.5 OpenRouter API
If you’re looking for the Xiaomi MiMo V2.5 OpenRouter model ID, it is:
xiaomi/mimo-v2.5
MiMo V2.5 is more interesting than the model ID alone suggests.
Xiaomi designed it as a native omni-modal model that can understand text, images, video, and audio, while also providing a 1M-token context window, tool calling, structured output, web search, and context caching.
For developers, that combination makes MiMo V2.5 particularly relevant for multimodal agents and long-context applications.
Xiaomi MiMo V2.5 OpenRouter Model ID
The OpenRouter model ID is:
xiaomi/mimo-v2.5
OpenRouter currently lists MiMo V2.5 with approximately 1.05 million tokens of context.
If you’re already using OpenRouter, a basic text request looks like this:
from openai import OpenAI
client = OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key="YOUR_OPENROUTER_API_KEY"
)
response = client.chat.completions.create(
model="xiaomi/mimo-v2.5",
messages=[
{
"role": "user",
"content": "Explain the architecture of this application."
}
]
)
print(response.choices[0].message.content)
The exact format for multimodal inputs depends on the API protocol and endpoint you’re using, so check the provider documentation before sending images, audio, or video.
Xiaomi MiMo V2.5 Pricing
Xiaomi also provides MiMo V2.5 through its own API.
Its current overseas pay-as-you-go pricing is:
| Usage | Price per 1M Tokens |
|---|---|
| Cached Input | $0.0028 |
| Uncached Input | $0.14 |
| Output | $0.28 |
Xiaomi lists a 1M-token context window, 128K maximum output, 100 RPM, and 10M TPM for its own MiMo V2.5 API.
These rate limits refer to Xiaomi’s API and should not automatically be assumed to apply to every third-party provider.
How Much Does MiMo V2.5 Cost on OpenRouter?
OpenRouter pricing can differ because multiple inference providers can serve the model.
Its main MiMo V2.5 model page currently shows a discounted headline price around:
$0.119 / 1M input tokens
$0.238 / 1M output tokens
However, OpenRouter’s provider pricing page shows different rates depending on the backend.
For example, the Xiaomi route has been listed at:
$0.14 / 1M input
$0.28 / 1M output
$0.0028 / 1M cache read
while other providers may charge more or less.
This is why API pricing should be compared using the actual provider route, not only the number shown in a search result.
Is Xiaomi MiMo V2.5 API Free?
The standard MiMo V2.5 API is not permanently free.
Xiaomi currently publishes usage-based pricing for the model, so production applications should assume normal token charges.
You may still encounter:
- Trial credits
- Temporary promotions
- Third-party discounts
- Developer campaigns
- Limited free quotas
But those should be treated as temporary access rather than permanent pricing.
This distinction is particularly important because Xiaomi currently lists some MiMo TTS models as temporarily free. That does not mean the main mimo-v2.5 language model is free.
So if you’re searching for a free Xiaomi MiMo API, the safest answer is:
MiMo V2.5 is a paid API model, although temporary free credits or promotional access may be available depending on the platform.
Why MiMo V2.5 Stands Out
MiMo V2.5 combines four features that don’t always appear together in one model:
- 1M-token context
- Native multimodal understanding
- Tool calling
- Very low official token pricing
Each of these matters for a different reason.
1M Context Window
Xiaomi officially lists:
Context Length: 1M tokens
Maximum Output: 128K tokens
A large context window can be useful for:
- Entire code repositories
- Large technical documents
- Long conversation history
- Agent memory
- Research collections
- Long video or multimodal context
The main benefit is not simply being able to send a huge prompt.
For agent applications, the context may gradually accumulate user instructions, intermediate results, tool calls, documents, and previous actions.
A larger context window gives the model more room to retain that state.
MiMo V2.5 Is Natively Multimodal
This is one of the biggest differences between MiMo V2.5 and many text-first LLMs.
Xiaomi lists the following input modalities:
Text
Image
Video
Audio
with text output.
That creates use cases such as:
- Screenshot understanding
- Image analysis
- Video summarization
- Audio understanding
- Visual document analysis
- Multimodal agents
For example, an agent could inspect an image, understand what it contains, reason about what should happen next, and then call a tool.
That is more useful than having separate disconnected models for every input type.
Does MiMo V2.5 Support Tool Calling?
Yes.
Xiaomi officially lists Tool Calls among MiMo V2.5’s capabilities. It also supports structured output, web search, streaming, and context caching.
A typical function schema might look like:
{
"type": "function",
"name": "get_order_status",
"description": "Get the current status of an order",
"parameters": {
"type": "object",
"properties": {
"order_id": {
"type": "string"
}
},
"required": ["order_id"]
}
}
The model can then decide that it needs this function, generate the appropriate arguments, and return a structured tool request.
Your application executes the actual function and returns the result to the model.
MiMo V2.5 for Multimodal Agents
The combination of multimodal understanding and tools is where MiMo V2.5 becomes especially interesting.
Consider a support agent.
A user uploads a screenshot showing an order problem and asks:
Why was my order cancelled?
The workflow could look like this:
User uploads screenshot
↓
MiMo V2.5 understands the screenshot
↓
Extracts order information
↓
Calls get_order_status()
↓
Receives live order data
↓
Explains the issue to the user
This is fundamentally different from a simple chatbot.
The model needs to perceive, reason, and act.
That is exactly the kind of workflow MiMo V2.5’s feature set is suited for.
MiMo V2.5 vs MiMo V2.5 Pro
Xiaomi currently offers both:
mimo-v2.5
mimo-v2.5-pro
The standard MiMo V2.5 is the native omni-modal model.
MiMo V2.5 Pro is positioned more aggressively toward complex, long-running agent tasks. Xiaomi describes the Pro model as capable of stable long-horizon workflows involving very large numbers of tool calls.
Pricing is also different:
| Model | Input | Output |
|---|---|---|
| MiMo V2.5 | $0.14/M | $0.28/M |
| MiMo V2.5 Pro | $0.435/M | $0.87/M |
Choose MiMo V2.5 when you care about:
- Multimodal input
- Low cost
- Long context
- General agent workflows
Choose MiMo V2.5 Pro when your priority is:
- More complex agent tasks
- Longer autonomous execution
- More demanding software-engineering workflows
Should You Use Xiaomi Direct or OpenRouter?
It depends on your architecture.
Xiaomi Direct API
Direct access makes sense if:
- MiMo is a core model in your stack
- You want Xiaomi’s native features
- You want direct billing
- You want the official pricing structure
OpenRouter
OpenRouter makes sense if:
- You’re already using OpenRouter
- You want MiMo alongside other OpenRouter models
- You prefer a single OpenRouter API integration
Unified Multi-Model Gateway
There is a third case:
You don’t necessarily want to build around Xiaomi or OpenRouter.
You want your own application to depend on one common API layer while the underlying model can change.
That is where a platform such as TokenHub fits.
Using TokenHub for Multi-Model Applications
TokenHub provides a RESTful API compatible with the OpenAI API format and exposes multiple major model providers behind one gateway.
The OpenAI-compatible Base URL is:
https://us-api.tokenhub.com/v1
A typical application can keep this structure:
from openai import OpenAI
client = OpenAI(
base_url="https://us-api.tokenhub.com/v1",
api_key="YOUR_TOKENHUB_API_KEY"
)
response = client.chat.completions.create(
model="MODEL_ID",
messages=[
{
"role": "user",
"content": "Analyze this task."
}
]
)
You then choose the model from TokenHub’s current model catalog.
This is useful when you want to compare MiMo’s capabilities with other models rather than designing your entire backend around a single provider.
Always check the live TokenHub catalog for current model availability and the exact model ID.
MiMo V2.5 vs Other Chinese LLMs
MiMo V2.5 competes in an increasingly crowded Chinese LLM market.
Depending on your workload, you may also want to test:
GLM
Strongly focused on long-horizon engineering and coding, with models offering 1M-class context.
Qwen
A broad model family covering coding, reasoning, agents, and multimodal workloads.
DeepSeek
Often evaluated for reasoning, coding, and price-sensitive API workloads.
MiniMax
Especially relevant for coding agents and long-running tool workflows.
The right choice depends on the actual workload.
A benchmark score doesn’t tell you whether a model will reliably call your own tools, understand your images, maintain context across your agent loop, or meet your latency target.
Test models with representative production tasks.
FAQ
What is the Xiaomi MiMo V2.5 OpenRouter model ID?
The model ID is:
xiaomi/mimo-v2.5
How much does MiMo V2.5 cost?
Xiaomi’s current overseas API pricing is $0.14/M uncached input tokens and $0.28/M output tokens. Cached input costs $0.0028/M.
Is MiMo V2.5 free?
The standard API is paid. Temporary trials or promotional credits may exist, but they are not the permanent price.
Does MiMo V2.5 support 1M context?
Yes. Xiaomi officially lists a 1M-token context window and 128K maximum output.
Does MiMo V2.5 support images?
Yes. It natively supports text, image, video, and audio input.
Does MiMo V2.5 support tool calling?
Yes. Tool calls are officially supported.
Build Without Locking Yourself to One Model
MiMo V2.5 is worth evaluating if you need long context, multimodal input, tool calling, and low token costs in the same model.
But your best model today may not be your best model six months from now.
TokenHub provides one OpenAI-compatible API gateway for supported AI models, making it easier to compare and switch models without rebuilding your integration around every new provider.