Guardrails lets you deploy large moderation models on dedicated computing resources. This topic describes the billing for this deployment mode.
Dedicated instance mode
In dedicated instance mode, the models used by Guardrails are deployed on dedicated computing resources. These resources are isolated from other customers' workloads. The dedicated models are not updated with the general-purpose edition.
The following services support this mode:
AI input content moderation (query_security_check)
AI-generated content moderation (response_security_check)
Billing for dedicated instances
The billing for dedicated instances includes computing resource fees and pay-as-you-go fees.
Computing resource fees:
Select a resource plan based on the length of the text that you submit for moderation:
LLM Dedicated Instance Resource Plan:
Specification: Supports a content length of up to 300 characters for moderation.
Purchase requirement: Requires a minimum of 20 queries per second (QPS) of computing resources. The list price is CNY 50,000 per month.
LLM Dedicated Instance Resource Plan Pro:
Specification: Supports a content length of up to 2,000 characters for moderation.
Purchase requirement: Requires a minimum of 20 QPS of computing resources. The list price is CNY 80,000 per month.
Pay-as-you-go fees: CNY 15 per 10,000 calls