Vertex AI vs. AWS Bedrock: the same model, two clouds
Claude on Vertex AI and Claude on AWS Bedrock are structurally parallel: both are cloud-native, fully-managed hosting surfaces for the identical underlying Claude models, and both differ from the direct Anthropic API in the same three ways -- cloud-native IAM auth instead of a portable API key, a cloud-specific model-ID addressing convention, and billing consolidated into the host cloud's own billing account rather than a separate Anthropic Console relationship.
| Vertex AI | AWS Bedrock | |
|---|---|---|
| Auth | Google Cloud IAM (service account / ADC / OAuth2) | AWS SigV4 signing, or bearer-token auth (C#/Go/Java SDKs only) |
| Model ID shape | publishers/anthropic/models/<model-id> | anthropic.<model-id>-v1:0 (ARN-style), or a global/regional prefix on newer models |
| Billing | Google Cloud billing account | AWS billing account |
| Capacity guarantee | GCP quota allocation / enterprise agreement | Provisioned Throughput (dedicated Model Units, hourly-billed) |
Neither surface changes the underlying model's capabilities or context window -- those are properties of the Claude model generation itself, identical everywhere it's served. This page is this site's own cross-reference rather than a restatement of subagentbedrock.com's content; see that site directly for the full, detailed Bedrock-side catalog (cross-region inference profiles, Guardrails, Provisioned Throughput specifics).
Sources: cloud.google.com/vertex-ai/.../use-claude; platform.claude.com/.../claude-on-amazon-bedrock.