This topic describes the features and pricing of the Moderation Agent.
Product features
Feature overview
Content Moderation lets you create and configure a custom Moderation Agent. This feature uses a large language model and flexible custom prompts to quickly detect and filter content that violates your business policies. The Moderation Agent currently supports the following modalities:
Text modality: The content to be detected is plain text.
Image modality: The content to be detected is a pure image.
Image-text multi-modality: A scenario involving a mix of image and text modalities.
Pricing
The Moderation Agent currently supports the pay-as-you-go billing method.
Pay-as-you-go
After you activate Content Moderation Enhanced Edition, the default billing method is pay-as-you-go. Charges apply only when you call the service. The billing logic for the Moderation Agent detection API is as follows:
Text modality: Billing is based on the text length and the total character count of the custom prompt.
Text length: Billed as one unit for every 2,000 characters. The API currently supports a maximum of 2,000 characters per request.
Custom prompt length: The first 3,000 characters are free. For prompts exceeding this limit, each 3,000-character increment, or part thereof, is billed as one additional unit.
Image modality: Billing is based on the number of images and the total character count of the custom prompt.
Number of images: Billed as one unit per image. The API currently supports only one image per request.
Custom prompt length: The first 3,000 characters are free. For prompts exceeding this limit, each 3,000-character increment, or part thereof, is billed as one additional unit.
Image-text multi-modality: Billing is based on the text length, the number of images, and the total character count of the custom prompt.
Text length: Billed as one unit for every 2,000 characters. The API currently supports a maximum of 2,000 characters per request.
Number of images: Billed as one unit per image. The API currently supports only one image per request.
Custom prompt length: The first 3,000 characters are free. For prompts exceeding this limit, each 3,000-character increment, or part thereof, is billed as one additional unit.
For example, after you enable the pay-as-you-go method, when you configure and call a Moderation Agent, your usage is metered and billed as shown in the following examples:
Scenario 1: Text modality. In this case, a custom prompt of 2,000 characters is metered once. This means that the call is metered once for the text modality, and the 2,000-character custom prompt (not exceeding 3,000) is free of charge.
Scenario 2: Image modality. If a custom prompt is configured with 4,000 characters, the call is metered twice. This consists of one time for the image call and one additional time for the 1,000 excess characters in the prompt.
Scenario 3: Image-text multi-modality. A call with a 4000-character custom prompt is metered 3 times: once for the text, once for the image, and once for the 1000-character excess in the custom prompt.
The following table lists the review types and their unit prices.
Modality | Review type | Unit price |
Text | Moderation Agent - Text - Per Call - Standard text_llm_standard | CNY 20 per 10,000 calls Note This item is billed when you select Text Moderation Large Model or Qwen3-Flash as the large model. |
Moderation Agent - Text - Per Call - Advanced text_llm_advanced | CNY 40 per 10,000 calls Note This item is billed when you select Qwen3-Plus as the large model. | |
Moderation Agent - Text - 3.6 - Per Call - Advanced text_llm36_tier_2 | CNY 100 per 10,000 calls Note This item is billed when you select Qwen3.6-Plus as the large model. | |
Image | Moderation Agent - Image - Per Call - Basic (image_vl_basic) | CNY 15 per 10,000 calls Note This item is billed when you select Qwen3-VL-Flash as the large model. |
Moderation Agent - Image - Per Call - Standard (image_vl_standard) | CNY 45 per 10,000 calls Note This item is billed when you select Qwen3-VL-Plus as the large model. | |
Image-text multi-modality | Moderation Agent - Image-text multi-modality - Per Call - Basic (image_vl_basic) | CNY 15 per 10,000 calls Note This item is billed when you select Qwen3-VL-Flash as the large model. |
Moderation Agent - Image-text multi-modality - Per Call - Standard (image_vl_standard) | CNY 45 per 10,000 calls Note This item is billed when you select Qwen3-VL-Plus as the large model. |
Usage
Integration
The Moderation Agent supports integration through the SDK or native API calls.
To integrate with the SDK, see Moderation Agent Service SDK Reference.
To integrate with native API calls, see Moderation Agent Synchronous Detection API Access Guide.
Console operations
You can configure a custom Moderation Agent in the Content Moderation console. For detailed instructions, see Build Custom Detection Capabilities with Moderation Agent.