Keep your OpenAI-compatible client
Change the base URL, API key, and model ID. Keep the request format you already use.
ONE OPENAI-COMPATIBLE API
Call leading AI models with one API key.
Use Chinese and global models through one endpoint, one USD balance, and a single usage history—without opening a separate account for every provider.
If a route fails, MaiToken switches to another eligible route automatically.
ROUTE + COST TRACE
See how one request becomes a USD charge.
The endpoint stays the same. The selected model, token counts, current input rate, and current output rate determine the call cost.
REQUESTPOST /v1/chat/completionsPrimary route selected
GATEWAYMaiToken · route + meterNo route switch needed
MODELdeepseek-v3 · example model ID
USAGE1,200 input · 380 output tokens
CHARGE BASISCurrent USD rate per 1M tokens
Request IDreq_mt_example_01
Automatic route switchingNo switch needed
Pricing unitCurrent USD rate per 1M tokens
MaiToken automatically switches to another eligible route. If no eligible route is available, the request may still fail.
Call cost formula
(input tokens ÷ 1,000,000 × current input rate) + (output tokens ÷ 1,000,000 × current output rate)
PRICE RELATIONSHIP
Most model rates are approximately 80%–90% of the corresponding public API list price.
See the Price Center for current rates.
One prepaid USD balance
Calls draw from one balance. Review model-level usage and charges in the console.
A Request ID for each call
Use it to find the route, status, token counts, cost, and fallback result.
Content and metadata have separate boundaries
MaiToken does not retain prompt or response content. Call metadata—including request/response token counts, model, route, status, and Request ID—is retained for billing and diagnostics.
01 / MODEL DIRECTORY
One directory for leading Chinese and global models.
Explore major model families through one account and API key. Compare capabilities here, then confirm current availability and rates before you build.
9 model families shown
In catalogOpenAI GPT
General chat, advanced reasoning, code, vision, and agent workflows.
TextVisionTool useJSON
In catalogAnthropic Claude
Long documents, careful analysis, nuanced writing, and complex coding tasks.
Long contextVisionTool use
In catalogGoogle Gemini
Multimodal understanding, long context, and high-throughput applications.
TextVisionAudio & videoTool use
In catalogDeepSeek
Reasoning, coding, and efficient batch workloads with strong support for Chinese-language tasks.
ChineseReasoningCoding
In catalogKimi
Long-context research, document analysis, search-assisted workflows, and tool use.
ChineseLong contextTool use
In catalogGLM
General-purpose options for Chinese-language, agent, and multimodal applications.
ChineseVisionTool use
In catalogQwen
Chinese-language understanding, structured output, code, and multimodal business workflows.
ChineseVisionTool use
In catalogMiniMax
Long-context text, speech, video, and multimodal generation for product use cases.
TextSpeechVideo
In catalogVolcengine / BytePlus
Text, vision, speech, and multimodal services for high-throughput applications.
TextVisionSpeech
Most model rates are approximately 80%–90% of the corresponding public API list price—roughly 10%–20% below that list price, not 80%–90% cheaper. Availability, token units, cached input, and non-text media pricing can vary by model. Confirm current availability and USD rates in the Model Directory / Price Center.
Open Model Directory / Price CenterFIT CHECK
No need to register multiple vendor accounts.
Tired of repeating setup and billing across different service providers? Sign up for MaiToken.
Provider-by-provider
With MaiToken
Accounts & keys
Create and maintain a separate account and key for each provider.
Use one account and one API key across models available to your account.
SDK setup
Adapt to provider-specific endpoints, SDKs, and model IDs.
Use one OpenAI-compatible endpoint and change the model ID.
Billing
Track separate balances, statements, and usage views.
Fund one prepaid USD balance and review unified usage.
Failure handling
Build and operate your own cross-route logic.
Automatically switch to another eligible route when a call fails or network conditions fluctuate.
FIRST CALL
01
02
03
From account to first paid call in three steps.
Registration does not create a key or include model credit.
Create your account
Register with email. Personal accounts do not require ID verification, a phone number, or a bank card.
Registration is freeCreate an API key
Open Console → API Keys → Create new key. MaiToken does not generate a key automatically after registration.
You control key creationAdd funds and make a request
Use a payment method offered at checkout, choose the model ID you want to call, and send your first call.
No free model-call credit
Example request
from openai import OpenAI
client = OpenAI(
base_url="https://api.maitoken.com/v1",
api_key="YOUR_MAITOKEN_API_KEY"
)
response = client.chat.completions.create(
model="deepseek-v3",
messages=[
{"role": "user", "content": "Compare two approaches to caching."}
]
)
print(response.choices[0].message.content)curl https://api.maitoken.com/v1/chat/completions \
-H "Authorization: Bearer YOUR_MAITOKEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v3",
"messages": [
{"role": "user", "content": "Compare two approaches to caching."}
]
}'import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.maitoken.com/v1",
apiKey: "YOUR_MAITOKEN_API_KEY"
});
const response = await client.chat.completions.create({
model: "deepseek-v3",
messages: [
{ role: "user", content: "Compare two approaches to caching." }
]
});
console.log(response.choices[0].message.content);Code could not be copied.
Select the code and copy it manually.
GATEWAY CONTROL
One endpoint. Automatic route switching built in.
When a call fails or network conditions fluctuate, MaiToken automatically switches to another eligible route to keep requests running and protect the user experience.
EXAMPLE REQUEST PATHIllustration · not live status
Your appOpenAI-compatible request
MaiToken gatewayRoute · switch · trace
Upstream model APISelected eligible route
OpenAI-compatible entry point
Keep a familiar request shape while changing the base URL, API key, and model ID.
Route-aware tracing
Use the Request ID to follow the selected route, status, and switching result.
Automatic route switching
The gateway selects another eligible route without requiring an endpoint change.
24/7 technical support
Share the Request ID so support can inspect call metadata and route behavior. Resolution times still depend on incident scope and upstream response.
DATA HANDLING
Know what passes through, what is retained, and who controls the next hop.
MaiToken routes and meters each model API call while the selected upstream provider remains responsible for its own policies.
Model request and response content
MaiToken processes request and response content long enough to forward the call between your application and the selected upstream model.
Content and call metadata
MaiToken does not retain prompt or response content. Call metadata is retained for billing and diagnostics.
PRICING, WITHOUT THE GUESSWORK
Fund once. Pay for measured usage. Review every call.
MaiToken uses a prepaid USD balance and per-use model rates.
Billing unit
Per 1M input and output tokens
Cached input, images, audio, or video may use separate units when a model supports them.
Settlement currency
USD
Prices, deposits, balance, usage, and charges are presented in USD on this page.
Payment methods
Shown at checkout
Use only a payment method the checkout page offers for your account and device.
Rate source
Price Center
Model rates can change. The current Price Center is the authoritative source for input, output, cached input, and other model-specific rates.
BEFORE YOU BUILD
Questions Before Your First Request
Straight answers about accounts, prices, model access, data handling, and failures.
How is pricing calculated?
Text-model rates are typically listed in USD per 1M input tokens and per 1M output tokens. Cached input and non-text media may use separate rates or units. Most model rates are approximately 80%–90% of the corresponding public API list price, which means roughly 10%–20% below that list price—not 80%–90% cheaper. Rates change, so use the Price Center as the authoritative source.
Which models can one key access?
One key can access models currently available to your account, including model families such as DeepSeek, Qwen, Kimi, GLM, MiniMax, and global options. Availability can depend on upstream policy, region, account controls, and platform risk rules. Confirm the current catalog in the console.
Does MaiToken store my prompt or response?
MaiToken does not retain prompt or response content. Call metadata—including request/response token counts, model, route, status, and Request ID—is retained for billing and diagnostics.
What happens when an upstream route fails?
When a call fails or network conditions fluctuate, MaiToken automatically switches to another eligible route to keep requests running and protect the user experience. If no eligible route is available, the request may still fail. Keep the Request ID and check the status page.
Is technical support available 24/7?
English and Chinese technical support is available 24/7. Response and resolution time depend on the incident scope, account context, and upstream provider response. Include the Request ID to speed diagnosis.
Is registration free? Are model calls free?
Registration is free. There is currently no free model-call credit or free model-usage tier. Add funds before making a live model call.
What is MaiToken?
MaiToken is an API gateway that lets one API key call multiple Chinese and global AI models. It provides an OpenAI-compatible entry point, unified metering, automatic route switching, Request IDs, and pay-as-you-go billing.
What do I need to create a personal account?
Personal registration does not require ID verification, a phone number, or a bank card. Registration does not automatically create an API key; create one in the console when you are ready.
Which currency will I actually be charged?
MaiToken presents deposits, balances, and usage charges in USD on this page and in the account flow. Checkout shows the final amount and available methods. Your bank may add fees, and a non-USD funding source may be converted at your issuer's rate. Applicable taxes or processor fees, if any, appear at checkout.
Which payment methods can I use?
Use a payment method offered on the actual checkout page. Available methods can depend on the processor, account, device, and transaction. MaiToken does not promise a specific card network, wallet, or bank method.
Can I use my existing OpenAI SDK?
Yes, for supported OpenAI-compatible calls. Replace the base URL, API key, and model ID. REST, Python, curl, and Node.js examples are available in the docs. Check model-specific differences before production.
Why not use each provider directly?
Direct access can be the simpler choice for one model or when you need provider-specific contracts and controls. MaiToken is useful when you want to compare or switch models without maintaining separate accounts, keys, interfaces, balances, and routing logic.
What about upstream retention or training?
Each upstream provider controls its own retention, training, regional, and legal policies. MaiToken cannot guarantee those policies on the provider's behalf. Review the selected provider's current terms before sending sensitive data.
What can support see when I report a failure?
Share the Request ID so technical support can inspect the call metadata retained for billing and diagnostics. Support cannot retrieve prompt or response content because MaiToken does not retain that content.
READY TO START?
Create the key. Choose the model. See the cost.
Use one OpenAI-compatible API and one USD balance across models available to your account.