AWS bedrock documentation change
Summary
Added Reserved tier documentation, expanded service tier details, and updated API parameters/quotas. Added regional availability table for Reserved tier.
Security assessment
The changes describe new service tiers (Reserved, Priority, Standard, Flex) and capacity management features. While Reserved tier mentions 99.5% uptime and overflow handling, there is no evidence of security vulnerability fixes or explicit security controls being added. The note about 1M context length unsupported in Reserved tier relates to functionality, not security.
Diff
diff --git a/bedrock/latest/userguide/service-tiers-inference.md b/bedrock/latest/userguide/service-tiers-inference.md index 41442ab74..2575c5482 100644 --- a//bedrock/latest/userguide/service-tiers-inference.md +++ b//bedrock/latest/userguide/service-tiers-inference.md @@ -4,0 +5,2 @@ +Reserved TierPriority TierStandard TierFlex TierUsing the service tier capability + @@ -7 +9,21 @@ -Amazon Bedrock offers three service tiers for model inference: Priority, Standard, and Flex. The Priority tier delivers the fastest response times for a price premium over standard on-demand pricing. It is best suited for mission-critical applications like customer-facing chatbots and real-time language translation services. The Standard tier provides consistent performance for everyday AI tasks, ideal for content generation, text analysis, and routine document processing. For workloads that can handle longer processing times, the Flex tier offers cost-effective processing for a pricing discount, perfect for model evaluations, content summarization, agentic workflows. +Amazon Bedrock offers four service tiers for model inference: Reserved, Priority, Standard, and Flex. With service tiers, you can optimize for availability, cost, and performance. + +## Reserved Tier + +The Reserved tier provides the ability to reserve prioritized compute capacity for your mission-critical applications that cannot tolerate any downtime. You have the flexibility to allocate different input and output tokens-per-minute capacities to match the exact requirements of your workload and control cost. When your application needs more tokens-per-minute capacity than what you reserved, the service automatically overflows to the Standard tier, ensuring uninterrupted operations. The Reserved tier targets 99.5% uptime for model response. Customers can reserve capacity for 1 month or 3 month duration. Customers pay a fixed price per 1K tokens-per-minute and are billed monthly. + +To get access to the Reserved tier, please contact your AWS account team. + +## Priority Tier + +The Priority tier delivers the fastest response times for a price premium over standard on-demand pricing. It is best suited for mission-critical applications with customer-facing business workflows that do not warrant 24X7 capacity reservation. Priority tier does not require prior reservation. You can simply set the "service_tier" optional parameter to "priority" to avail request level prioritization. Priority tier requests are prioritized over Standard and Flex tier requests. + +## Standard Tier + +The Standard tier provides consistent performance for everyday AI tasks such as content generation, text analysis, and routine document processing. By default all inference requests are routed to the Standard tier when the "service_tier" parameter is missing. You can also set the "service_tier" optional parameter to "default" for your inference request to be served with Standard tier. + +## Flex Tier + +For workloads that can handle longer processing times, the Flex tier offers cost-effective processing for a pricing discount. This helps you optimize cost for workloads such as model evaluations, content summarization, and agentic workflows. You can set the "service_tier" optional parameter to "flex" for your inference request to be served with the Flex tier and avail the pricing discount. + +## Using the service tier capability @@ -9 +31 @@ Amazon Bedrock offers three service tiers for model inference: Priority, Standar -Accessing the service tier capability requires no additional setup, allowing for immediate enhancement of existing applications with price/performance optimization. You can set the "service_tier" optional parameter to "priority", "default", or "flex" while calling the Amazon Bedrock runtime API. If you select "default" as your invocation option, your requests will be served by standard inference. By default all requests are routed to through standard inference. +To access the service tier capability, you can set the "service_tier" optional parameter to "reserved", "priority", "default", or "flex" while calling the Amazon Bedrock runtime API. @@ -12 +34 @@ Accessing the service tier capability requires no additional setup, allowing for - "service_tier" : "priority | default | flex" + "service_tier" : "reserved | priority | default | flex" @@ -14 +36 @@ Accessing the service tier capability requires no additional setup, allowing for -Your on-demand quota for a model is shared across all service tiers. The service tier configuration for a served request is visible in API response and AWS CloudTrail Events. You can also view metrics for service tiers in Amazon CloudWatch Metrics under "model-id+priority" and "model-id+flex". +Your on-demand quota for a model is shared across the "priority", "default", and "flex" service tiers. Your "reserved" tier capacity reservation is separate from your on-demand quota. The service tier configuration for a served request is visible in API response and AWS CloudTrail Events. You can also view service tier metrics in Amazon CloudWatch Metrics under ModelId, ServiceTier, and ResolvedServiceTier, where ResolvedServiceTier shows the actual tier that served your requests. @@ -17,0 +40,32 @@ For more information about pricing, visit the [pricing page](https://aws.amazon. +Models and regions supported by the Reserved service tier: + +**Provider** | **Model** | **Model IDs** | **Regions** +---|---|---|--- +Anthropic | Claude Sonnet 4.5 | global.anthropic.claude-sonnet-4-5-20250929-v1:0us.anthropic.claude-sonnet-4-5-20250929-v1:0 | ap-northeast-1 +ap-northeast-2 +ap-northeast-3 +ap-southeast-1 +ap-southeast-2 +ap-south-1 +ap-southeast-3 +ap-south-2 +ap-southeast-4 +ca-central-1 +Europe-west-1 +Europe-central-1 +Europe-central-2 +Europe-north-1 +Europe-south-1 +Europe-south-2 +Europe-west-2 +Europe-west-3 +sa-east-1 +us-east-1 +us-east-2 +us-west-1 +us-west-2 + +###### Note + +1M context length for Sonnet 4.5 is not supported by the Reserved tier. +