tencent cloud

Processing Fee

Download
Focus Mode
Font Size
Last updated: 2026-11-17 14:27:52
AI-Translated & Reviewed
The processing fee is charged for a Cloud Model Routing (CMR) instance in pay-as-you-go billing mode. After an instance is created successfully and used for business processing, corresponding fees are generated.

Billing Cycle

Billing is based on the actual number of tokens processed by a CMR instance, and settlement is performed once every hour.
If you do not use the instance at all after creating it, no fees are generated.
The processing fee charged by a CMR instance does not vary with the unit price of the upstream model. It is positively correlated only with the total number of tokens processed via the gateway.
The processing fee generated by a CMR instance can only be deducted by using a resource package. For details, see Resource Package Introduction.
Attention:
You need to purchase a resource package before you can create a pay-as-you-go CMR instance.

Pricing Details

Region
Unit Price (USD/Million Tokens)
Guangzhou/Shanghai/Beijing/Singapore/Silicon Valley
0.07

Billing Formula

The formula for calculating the processing fee of a pay-as-you-go instance is as follows:
CMR processing fee = Unit price of processing fee (USD/million tokens) x Total number of processed tokens (in millions of tokens)
The total number of processed tokens is approximately equal to total_tokens obtained from usage returned by the upstream model. This metric refers to the sum of all input tokens and output tokens routed via the gateway in a single request.
Note:
total_tokens consists of input and output tokens. Both are included in the processing fee:
Input tokens (prompt_tokens): the sum of the detailed items from usage.prompt_tokens_details, including:
Non-cached input tokens
Input tokens read from the cache
Input tokens written to the cache
Output tokens (completion_tokens): the sum of the detailed items from usage.completion_tokens_details, including:
Regular response tokens (completion_tokens)
Reasoning/thinking tokens of the model (reasoning_tokens if any)
Additional tokens generated for completing tool calls

Billing Example

Scenario: You create a pay-as-you-go CMR instance in the Guangzhou region. The instance is created successfully at 00:00:00 on July 1, 2026. At 02:30:00 on July 1, 2026, assume that you call an upstream model once via our gateway (using Kimi as an example).
Example of a user request and an upstream model response:

{
"id": "cmpl-a1b2c3d4e5f6a7b8c9d0e1f2",
"object": "chat.completion",
"created": 1774675200,
"model": "kimi-k2-0905-preview",
"choices": [
{
"index": 0,
"finish_reason": "stop",
"message": {
"role": "assistant",
"content": "Hello"
},
"logprobs": null
}
],
"usage": {
"prompt_tokens": 98,
"completion_tokens": 12,
"total_tokens": 110,
"prompt_tokens_details": {
"cached_tokens": 64 // ← The number of cache-hit tokens. This field is automatically populated and is a subset of prompt_tokens.
},
"completion_tokens_details": {
"reasoning_tokens": 0 // ← This value is non-zero for thinking models, and the unit price is the same as that of regular output.
}
}
}



The processing fee for a CMR instance is calculated as follows:

1. Determine the total number of processed tokens:
The gateway extracts the usage.total_tokens field from the response and obtains a value of 110 tokens.
2. Calculate the token processing fee:
Total number of processed tokens = 110 tokens
Unit price of the processing fee = USD 0.07/million tokens
CMR processing fee = 110/1,000,000 x 0.07 ≈ USD 0.0000077
3. For this call, the CMR processing fee generated for you is approximately USD 0.0000077. The fee remains fixed regardless of whether the upstream model for this call is Kimi, GLM, or DeepSeek, as long as the total number of processed tokens is 110.

References

Help and Support

Was this page helpful?

Help us improve! Rate your documentation experience in 5 mins.

Feedback