Pay-as-you-go-AI Guardrails(AI Guardrails)-阿里云帮助中心

This topic describes how pay-as-you-go billing works for AI Guardrails, including metering rules, the review types and unit prices for content moderation and Detection Agent, and billing, overdue payment, and cancellation details.

Pay-as-you-go is the default billing method for AI Guardrails. It suits scenarios where usage fluctuates significantly and you want to pay only for the content that you check. Activation is free of charge, and the system bills you based on actual usage. Charges are metered by the content checked and the volume of checks, and are deducted from your Alibaba Cloud account each billing cycle.

Important

For discounts, purchase a savings plan or a resource plan. For details, see Savings plans and resource plans.
If the overdue payment lasts more than 7 days, the instance is released. Settle the overdue amount and activate the service again.
Prompt injection moderation has been commercially billed since January 22, 2026. For details, see the notes in the unit price tables below.

Billing overview

The billing cycle window for AI Guardrails is 5 minutes.
In each billing cycle window, Alibaba Cloud generates a bill based on your usage in the previous window and deducts the service fee from your Alibaba Cloud account according to the bill amount.
Pay-as-you-go is billed by the content checked and the volume of checks. The metering unit differs by access method and review type. For unit prices, see the following sections.
If you have received an AI Guardrails trial package (free quota) or purchased a resource plan, the system deducts from it first. Only usage that exceeds the quota or occurs after it expires is billed at pay-as-you-go rates.

Unit prices

Unit prices for AI Guardrails vary by service. Each modality has its corresponding services, and different services support different protection dimensions. Each protection dimension is billed separately and is not billed unless enabled. For the services supported by AI Guardrails, see Supported moderation services.

Content compliance
Sensitive content
Prompt injection
Custom detection agents
Well-known IP detection agents
Malicious URL detection
Model hallucination detection
Malicious file detection
Digital watermarking

The following sections describe the unit price of each protection dimension. The metering units used in the unit prices have the following meanings:

"Per call" or "per 10,000 calls" is metered by the number of calls.
"Slices" are metered by fixed-length character slices, and a partial slice counts as one slice. The slice size is 3,000 characters for Detection Agent and 2,000 characters for synchronous file moderation.
Audio is metered by minute.
Dedicated services used by Model Studio integration (to be discontinued) are metered by token.

Content compliance

The following are the pay-as-you-go unit prices for content compliance moderation, billed by the number of calls or the number of slices. For the access method supported by each modality, see the scope notes at the end of this topic. Rows with the same modality and review type are merged, and the supported services are listed on separate lines in the cell.

Modality	Region	Review type (moderationType)	Services (Service)	Unit price
Text	China/Cross-border	`text_guard_advanced`	AI input content security check (Pro) (`query_security_check_pro`) AI-generated content security check (Pro) (`response_security_check_pro`) AI input content security check (`query_security_check`) AI-generated content security check (`response_security_check`) AI input content security check - cross-border edition (`query_security_check_cb`) AI-generated content security check - cross-border edition (`response_security_check_cb`)	CNY 15 per 10,000 calls
Image	China	`image_guard_advanced`	AIGC input image security check (`img_query_security_check`) AIGC output image security check (`img_response_security_check`)	CNY 30 per 10,000 calls
Image	Cross-border	`image_guard_advanced_cb`	AIGC input image security check (Pro) - cross-border edition (`img_query_guard_pro_cb`) AIGC output image security check (Pro) - cross-border edition (`img_response_guard_pro_cb`)	CNY 75 per 10,000 calls
Multimodal (text and image)	China	`text_image_guard_advanced`	Multimodal (text and image) content security check (`text_img_security_check`)	CNY 45 per 10,000 calls
File	China/Cross-border	`file_guard_sync_advanced`	Real-time file check (`file_security_sync_check`), real-time file check (`file_sync_check_cb`)	CNY 20 per 10,000 slices Note Text in the file is sliced by 2,000 characters, and the system meters by the number of slices.
Multimodal (text and file)	China	`file_guard_sync_advanced`	Real-time multimodal (text file) check (`text_file_sec_sync_check`)	CNY 20 per 10,000 slices Note Text in the file is sliced by 2,000 characters, and the system meters by the number of slices.
Audio	China	`audio_guard_standard`	AIGC audio security check (`audio_security_check`)	CNY 300 per 10,000 minutes Note Audio moderation is metered by audio duration in minutes, regardless of how many detection features are enabled.
Audio	Cross-border	`audio_guard_standard_cb`	AIGC audio security check (`audio_security_check_cb`)	CNY 750 per 10,000 minutes Note Audio moderation is metered by audio duration in minutes, regardless of how many detection features are enabled.
Video	China	`image_guard_advanced`, `audio_guard_standard`	AIGC video security check (`video_security_check`)	Image cost: CNY 30 per 10,000 frames Audio cost: CNY 300 per 10,000 minutes Note The video moderation fee is the sum of the Image cost and the Audio cost: Frame moderation is metered by the actual number of captured frames. The number of captured frames is determined by the Frame Capture Method and the corresponding Frame Capture Frequency or Total Frames. Audio in the video is moderated only after Enable Audio Moderation is enabled.
Video	Cross-border	`image_guard_advanced_cb`, `audio_guard_standard_cb`	AIGC video security check (`video_security_check_cb`)	Image cost: CNY 75 per 10,000 frames Audio cost: CNY 750 per 10,000 minutes Note The video moderation fee is the sum of the Image cost and the Audio cost: Frame moderation is metered by the actual number of captured frames. The number of captured frames is determined by the Frame Capture Method and the corresponding Frame Capture Frequency or Total Frames. Audio in the video is moderated only after Enable Audio Moderation is enabled.

Modality	Region	Review type (moderationType)	Services (Service)	Unit price
Text	International	`text_guard_advanced`	AI input content security check (`query_security_check_intl`) AI-generated content security check (`response_security_check_intl`)	USD 6 per 10,000 calls
Text	China (inbound)	`text_guard_advanced_cn`	AI input content security check (Pro) (`query_guard_pro_ec`) AI-generated content security check (Pro) (`response_guard_pro_ec`)	USD 2.4 per 10,000 calls
Image	International	`image_guard_advanced`	AIGC input image security check (`img_query_guard_intl`) AIGC output image security check (`img_response_guard_intl`)	USD 12 per 10,000 calls
Image	China (inbound)	`image_guard_advanced_cn`	AIGC input image security check (`img_query_guard_ec`) AIGC output image security check (`img_response_guard_ec`)	USD 4.7 per 10,000 calls
Multimodal (text and image)	International	`text_image_guard_advanced`	Multimodal (text and image) content security check (`text_img_security_check_intl`)	USD 18 per 10,000 calls
File	International	`file_guard_sync_advanced`	Real-time file check (`file_sync_check_intl`)	USD 8 per 10,000 slices Note Text in the file is sliced by 2,000 characters, and the system meters by the number of slices.
Audio	International	`audio_guard_standard`	AIGC audio security check (`audio_security_check_intl`)	USD 120 per 10,000 minutes Note Audio moderation is metered by audio duration in minutes, regardless of how many detection features are enabled.
Audio	China (inbound)	`audio_guard_standard_cn`	AIGC audio security check (`audio_security_check_ec`)	USD 4.7 per 10,000 minutes Note Audio moderation is metered by audio duration in minutes, regardless of how many detection features are enabled.
Video	International	`image_guard_advanced`, `audio_guard_standard`	AIGC video security check (`video_security_check_intl`)	Image cost: USD 6 per 10,000 frames Audio cost: USD 120 per 10,000 minutes Note The video moderation fee is the sum of the Image cost and the Audio cost: Frame moderation is metered by the actual number of captured frames. The number of captured frames is determined by the Frame Capture Method and the corresponding Frame Capture Frequency or Total Frames. Audio in the video is moderated only after Enable Audio Moderation is enabled.

Sensitive content

Sensitive content moderation is billed independently of content compliance moderation. It supports API or AI Gateway access and is billed by the number of calls.

Modality	Region	Review type (moderationType)	Services (Service)	Unit price
Text, image, file	China	`text_sddp_advanced`	AI input content security check (Pro) (`query_security_check_pro`) AI-generated content security check (Pro) (`response_security_check_pro`) AI input content security check (`query_security_check`) AI-generated content security check (`response_security_check`) AI input content security check - cross-border edition (`query_security_check_cb`) AI-generated content security check - cross-border edition (`response_security_check_cb`) AIGC input image security check (`img_query_security_check`) AIGC output image security check (`img_response_security_check`) Multimodal (text and image) content security check (`text_img_security_check`) Real-time file check (`file_security_sync_check`), real-time file check (`file_sync_check_cb`) Real-time multimodal (text file) check (`text_file_sec_sync_check`)	CNY 15 per 10,000 calls
Text, image	Cross-border	`text_sddp_advanced_cb`	AIGC input image security check (Pro) - cross-border edition (`img_query_guard_pro_cb`) AIGC output image security check (Pro) - cross-border edition (`img_response_guard_pro_cb`)	CNY 38 per 10,000 calls

Prompt injection

Prompt injection detection is billed independently. It supports API or AI Gateway access and is billed by the number of calls.

Modality	Region	Review type (moderationType)	Services (Service)	Unit price
Text, image, file	China	`text_guard_prompt_attack`	AI input content security check (Pro) (`query_security_check_pro`) AI-generated content security check (Pro) (`response_security_check_pro`) AI input content security check (`query_security_check`) AI-generated content security check (`response_security_check`) AI input content security check - cross-border edition (`query_security_check_cb`) AI-generated content security check - cross-border edition (`response_security_check_cb`) AIGC input image security check (`img_query_security_check`) AIGC output image security check (`img_response_security_check`) Multimodal (text and image) content security check (`text_img_security_check`) Real-time file check (`file_security_sync_check`), real-time file check (`file_sync_check_cb`) Real-time multimodal (text file) check (`text_file_sec_sync_check`)	CNY 15 per 10,000 calls
Text, image	Cross-border	`text_guard_prompt_attack_cb`	AIGC input image security check (Pro) - cross-border edition (`img_query_guard_pro_cb`) AIGC output image security check (Pro) - cross-border edition (`img_response_guard_pro_cb`)	CNY 38 per 10,000 calls

Custom detection Agent/Well-known IP detection Agent

Billing depends on the total number of prompt characters and the selected model. The billing rules are as follows:

Total custom prompt characters: Prompts are sliced by 3,000 characters, and each 3,000 characters is billed once (fewer than 3,000 characters are billed as 3,000). For example, if the total custom prompt is 4,000 characters, it is sliced into 2 slices and billed 2 times.
Selected model: Different models correspond to different billable items. For the mapping between models and unit prices, see Detection Agent unit price table. For example, for Qwen3Guard-Gen-4B, the prompt is built in, so billing is independent of prompt length and based only on the number of calls.

Modality	Review type (moderationType)	Model	Services (service)	Unit price
Text	Detection Agent - Basic text_guard_llm_standard	Qwen3.0-Flash, Turbo Edition	AI input content security check - Pro (query_security_check_pro) AI-generated content security check - Pro (response_security_check_pro) AI input content security check (query_security_check) AI-generated content security check (response_security_check)	CNY 20 per 10,000 slices
Text	Detection Agent - Advanced text_guard_llm_advanced	Qwen3.0-Plus	AI input content security check - Pro (query_security_check_pro) AI-generated content security check - Pro (response_security_check_pro) AI input content security check (query_security_check) AI-generated content security check (response_security_check)	CNY 40 per 10,000 slices
Text	Detection Agent - LLM36 - Level 1 text_guard_llm36_tier_1	Qwen3.6-Flash	AI input content security check - Pro (query_security_check_pro) AI-generated content security check - Pro (response_security_check_pro) AI input content security check (query_security_check) AI-generated content security check (response_security_check)	CNY 60 per 10,000 slices
Text	Detection Agent - LLM36 - Level 2 text_guard_llm36_tier_2	Qwen3.6-Plus, Advanced Edition	AI input content security check - Pro (query_security_check_pro) AI-generated content security check - Pro (response_security_check_pro) AI input content security check (query_security_check) AI-generated content security check (response_security_check)	CNY 100 per 10,000 slices
Text	Detection Agent - Basic - cross-border edition text_guard_llm_standard_cb	Qwen3.0-Flash, Turbo Edition	AI input content security check - cross-border edition (query_security_check_cb) AI-generated content security check - cross-border edition (response_security_check_cb)	CNY 50 per 10,000 slices
Text	Detection Agent - Advanced - cross-border edition text_guard_llm_advanced_cb	Qwen3.0-Plus	AI input content security check - cross-border edition (query_security_check_cb) AI-generated content security check - cross-border edition (response_security_check_cb)	CNY 100 per 10,000 slices
Text	Detection Agent - LLM36 - Level 1 - cross-border edition text_guard_llm36_tier_1_cb	Qwen3.6-Flash	AI input content security check - cross-border edition (query_security_check_cb) AI-generated content security check - cross-border edition (response_security_check_cb)	CNY 150 per 10,000 slices
Text	Detection Agent - LLM36 - Level 2 - cross-border edition text_guard_llm36_tier_2_cb	Qwen3.6-Plus, Advanced Edition	AI input content security check - cross-border edition (query_security_check_cb) AI-generated content security check - cross-border edition (response_security_check_cb)	CNY 250 per 10,000 slices
Image	Detection Agent - Image - Basic image_guard_vl_basic	Turbo Edition	AIGC input image security check (img_query_security_check) AIGC output image security check (img_response_security_check)	CNY 15 per 10,000 slices
Image	Detection Agent - LLM36 - Level 2 text_guard_llm36_tier_2	Advanced Edition	AIGC input image security check (img_query_security_check) AIGC output image security check (img_response_security_check)	CNY 100 per 10,000 slices
Image	Detection Agent - Image - Basic - cross-border edition image_guard_vl_basic_cb	Turbo Edition	AIGC input image security check (img_query_security_check) AIGC output image security check (img_response_security_check)	CNY 38 per 10,000 slices
Image	Detection Agent - LLM36 - Level 2 - cross-border edition text_guard_llm36_tier_2_cb	Advanced Edition	AIGC input image security check (img_query_security_check) AIGC output image security check (img_response_security_check)	CNY 250 per 10,000 slices

Features in public preview

The following detection dimensions are in public preview and free of charge:

Protection dimension	Description
Malicious URL detection	Scans LLM content for phishing, trojan, and non-compliant URLs.
Model hallucination detection	Identifies false or inaccurate information generated by LLMs based on knowledge base consistency checks.
Malicious file detection	Detects executable scripts, macro viruses, and nested attack code in PDF, PPT, and DOC files.
Digital watermarking	Embeds visible watermarks in AI-generated images for traceability and copyright protection.

Model Studio integration unit prices (to be discontinued)

Important

The integration of AI Guardrails with Model Studio has been upgraded. The new solution no longer supports Model Studio-specific services and supports only general services, so that protection capabilities stay in sync. Therefore, the Model Studio-specific services listed below will be discontinued and can no longer be added. For new integrations, see the new integration solution.

Model Studio integration is billed by token. A single check that uses fewer than 1,000 tokens is billed as 1,000 tokens, and a check that uses more than 1,000 tokens is billed by actual token count.

Review type	Supported business scenarios (service)	Unit price
Text compliance moderation - per token - Advanced text_guard_token_advanced	Model Studio input content security guardrail - Qwen3Guard edition - Pro (qwen_query_check_pro) Model Studio output content security guardrail - Qwen3Guard edition - Pro (qwen_response_check_pro) Model Studio input content security guardrail - Pro (bl_query_guard_pro) Model Studio output content security guardrail - Pro (bl_response_guard_pro)	CNY 0.003 per 1,000 tokens
Text compliance moderation - per token - Basic text_guard_token_standard	Model Studio input content security guardrail - Qwen3Guard edition (qwen_query_check) Model Studio output content security guardrail - Qwen3Guard edition (qwen_response_check) Model Studio input content security guardrail (bl_query_guard) Model Studio output content security guardrail (bl_response_guard)	CNY 0.0004 per 1,000 tokens
Sensitive content moderation - per token - Advanced text_sddp_token_advanced	Model Studio input content security guardrail - Qwen3Guard edition - Pro (qwen_query_check_pro) Model Studio output content security guardrail - Qwen3Guard edition - Pro (qwen_response_check_pro) Model Studio input content security guardrail - Pro (bl_query_guard_pro) Model Studio output content security guardrail - Pro (bl_response_guard_pro)	CNY 0.003 per 1,000 tokens
Sensitive content moderation - per token - Basic text_sddp_token_standard	Model Studio input content security guardrail - Qwen3Guard edition (qwen_query_check) Model Studio output content security guardrail - Qwen3Guard edition (qwen_response_check) Model Studio input content security guardrail (bl_query_guard) Model Studio output content security guardrail (bl_response_guard)	CNY 0.0004 per 1,000 tokens
Prompt injection moderation	Model Studio input content security guardrail - Qwen3Guard edition - Pro (qwen_query_check_pro) Model Studio output content security guardrail - Qwen3Guard edition - Pro (qwen_response_check_pro) Model Studio input content security guardrail - Qwen3Guard edition (qwen_query_check) Model Studio output content security guardrail - Qwen3Guard edition (qwen_response_check) Model Studio input content security guardrail - Pro (bl_query_guard_pro) Model Studio output content security guardrail - Pro (bl_response_guard_pro) Model Studio input content security guardrail (bl_query_guard) Model Studio output content security guardrail (bl_response_guard)	CNY 15 per 10,000 calls Important This feature has been commercially billed since January 22, 2026. If you do not need it, disable the service in the AI Guardrails console to avoid charges.
Image moderation - per call - Advanced image_guard_advanced	Model Studio input image security guardrail (bl_img_query_guard) Model Studio output image security guardrail (bl_img_response_guard)	CNY 30 per 10,000 calls

Overdue payments and top-ups

Deduction account and associated services

AI Guardrails charges are deducted directly from the balance of your Alibaba Cloud primary account, which is shared with Model Studio. There is no separate Model Studio balance. Activating AI Guardrails does not activate monitoring services such as ARMS or OpenTelemetry. These services must be activated separately, and no related charges are incurred if they are not activated.

Overdue payments and service suspension

When the system detects that your account is in arrears, the service is suspended. After you pay the outstanding balance, you can use AI Guardrails again (effective within 5 minutes). For details about how to top up an account in arrears, see Top-up instructions. If a pay-as-you-go account remains in arrears for too long (more than 7 days), the instance is released. You must pay the outstanding balance and activate the service again before you can use it.

FAQ

Does the AI Guardrails bill support viewing cost details by instance ID or service type? What is the UID in an instance ID?

AI Guardrails is billed centrally by Alibaba Cloud account, and cost details cannot be split by instance ID or by specific service type (such as content moderation). To view your bill, go to Cost Center > Bills.

The instance ID in an exported bill is in the format 163245*****00260;text_guard_advanced, where the first part (such as 163245*****00260) is the Alibaba Cloud account UID (the unique identifier of the account) and the second part (such as text_guard_advanced) is the service type identifier. The two parts are separated by a semicolon (;).

How do I cancel an order, unsubscribe, release resources, or disable the service to stop billing?

The pay-as-you-go AI Guardrails service cannot be disabled, unsubscribed, or canceled directly, and activation itself is free of charge. To stop billing, stop calling all moderation API operations in your business code. No new charges are incurred after you stop calling, and you can confirm in subsequent bills that the service is no longer charged.

The bill has been charged. Why wasn't the resource plan fully deducted?

Go to Cost Center > Bill Details to view the deduction details. A resource plan covers only specific moderation types or quotas, and usage beyond that is billed at pay-as-you-go rates. The bill may also include other billable items that the resource plan does not cover.