AWS Security ChangesHomeSearch

AWS wellarchitected documentation change

Service: wellarchitected · 2025-11-22 · Documentation low

File: wellarchitected/latest/generative-ai-lens/gencost02-bp01.md

Summary

Updated cost optimization guidance: Added scenario comparing SageMaker vs Bedrock costs, refined implementation steps, and updated section headers.

Security assessment

Changes focus exclusively on cost/performance tradeoffs in model hosting. No security controls, vulnerabilities, or security features are mentioned in the added scenario or modified steps.

Diff

diff --git a/wellarchitected/latest/generative-ai-lens/gencost02-bp01.md b/wellarchitected/latest/generative-ai-lens/gencost02-bp01.md
index 37b8377dd..f13740679 100644
--- a//wellarchitected/latest/generative-ai-lens/gencost02-bp01.md
+++ b//wellarchitected/latest/generative-ai-lens/gencost02-bp01.md
@@ -27,0 +28,2 @@ Throughput sensitive workloads often require additional resources to service inf
+Consider a scenario where you want to serve inference capabilities for a single model for small, periodic workloads. Evaluate the cost of hosting this model on an Amazon SageMaker AI inference endpoint. Compare these costs against the cost of importing the model to Amazon Bedrock using Amazon Bedrock's Custom Model Import feature and using API-based inference. Evaluate the cost to deploy this model using either paradigm and compare them with respect to the total cost of ownership. Where performance trade-offs are negligible, deploy to the most cost-effective inference paradigm. 
+
@@ -34 +36,3 @@ Throughput sensitive workloads often require additional resources to service inf
-  3. Select the most appropriate, lowest-cost hosting option. 
+  3. Select and test the available optinos that satisfy the workload requirements for latency, throughput, and response quality. 
+
+  4. Implement the most appropriate, lower-cost hosting option for your model serving paradigm (for example, managed or self-hosted). 
@@ -41 +45 @@ Throughput sensitive workloads often require additional resources to service inf
-**Related practices:**
+**Related best practices:**
@@ -52 +56 @@ Throughput sensitive workloads often require additional resources to service inf
-**Related guides, videos, and documentation:**
+**Related documents:**