Integrations
AWS adds model fallback to AgentCore Gateway
Vendor-reportedChecked: [release]
Editorial attribution: Gosia Madej
Prepared with AI assistance. Editorial attribution does not imply personal product testing or manual review of every story.
Published: · Event date:
What changes for users
For support agents, resilience and answer consistency need separate acceptance checks. Test every allowed destination against the same support policy and tool contracts before enabling it. Include failure attempts in a cost estimate and correlate routing logs with customer-facing answers. A successful HTTP response alone cannot establish that a substitute model followed the same policy.
Editorial assessmentChecked: [guide]

AWS says InvokeModel and InvokeModelWithResponseStream support and model fallback are available in Gateway regions except China. Rules specify alternative destinations and failure triggers.
Vendor-reportedChecked: [release]
The guide allows one or two alternatives. Once streaming starts, fallback stops. Routing logs identify the serving destination; the caller receives no destination indicator. Each outbound attempt incurs a gateway invocation charge, with model costs billed separately.
Vendor-reportedChecked: [guide]
Checked against sources
AI-assisted source check
How this was checked
Source evidence
Original source excerpt. This shows what the source says; it does not establish product performance.
Each outbound attempt the gateway makes counts as a gateway invocation
Sources
- [release] AWS: October 22 InvokeModel and fallback announcementVendor-reportedPublished: 22 October 2026 · Checked: 24 October 2026
- [guide] AWS: model routing, streaming limits, observability and billingVendor-reportedSource publication date not stated · Checked: 24 October 2026