The AWS Console Works. Why Does Your Bedrock Code Fail? #aws #amazonbedrock #genai

Invocation of model ID anthropic.claude-sonnet-5-5 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model. Fixing the AWS Bedrock ValidationException takes two steps, in order: call the model through an inference profile (global., us., eu., au.), then grant IAM on the profile and on the foundation model in every Region the profile routes to. Tested with the AWS CLI, Python (boto3) and the AWS SDK for JavaScript: same model ID in, same error out. CHAPTERS 00:00 Console works, code fails 01:18 What the error actually means 02:50 Why the console works 03:14 The advice that's half right 04:36 Error #2: AccessDenied 05:49 The IAM statement that fixes it 06:45 Global vs geo: the 10% difference 07:25 Find the profile ID: list-inference-profiles 07:38 Reference card REFERENCE CARD bare model ID → ValidationException profile ID (global. / us. / eu.) → AccessDenied? → IAM: inference-profile + foundation-model in every destination Region global. = list price - us. / eu. = +10% PRICES USED (est., Standard tier, checked 2026-10-12) Claude Sonnet 5.5 on Amazon Bedrock: Global profile $2 / $10 per 1M input / output tokens; Geo profiles (us., eu., au.) +10% ($2.20 / $11). Example: 100M input + 20M output per month = $400 Global vs $440 Geo. SOURCES Claude Sonnet 5.5 model card: https://docs.aws.amazon.com/bedrock/latest... Inference profile prerequisites (IAM): https://docs.aws.amazon.com/bedrock/latest... Inference profiles (system-defined and application): https://docs.aws.amazon.com/bedrock/latest... Amazon Bedrock pricing: https://aws.amazon.com/bedrock/pricing/ #AWS #AmazonBedrock #GenAI