Skip to main content
resource · aws

Amazon Bedrock Marketplace Endpoint

live rule families
1
schedulable
no
category
ai-ml-services

Does ZopNight manage Amazon Bedrock Marketplace Endpoint?

Bedrock Marketplace endpoints deploy third-party models onto dedicated instances that bill per instance-hour, plus any software fee the vendor sets, and traffic never enters the equation. ZopNight discovers endpoints via its Bedrock provider on the 6-hour cycle, attributes cost from Cost Explorer or CUR 2.0, and flags idle endpoints.

Rules that fire on Amazon Bedrock Marketplace Endpoint

At a glance

Amazon Bedrock Marketplace Endpoint coverage facts.
Field Value
Scheduling notesdiscovery, cost tracking, and recommendations only.

A Bedrock Marketplace endpoint deploys a third-party model from the marketplace onto dedicated instances, billed per instance-hour plus any software fees. Like SageMaker endpoints, marketplace endpoints bill continuously regardless of traffic.

Instance-hours with a vendor margin

Marketplace models step outside Bedrock’s serverless pricing entirely. Where the catalog’s first-party models bill per token on demand, a marketplace deployment provisions dedicated instances that bill per instance-hour from deployment to deletion, and the model vendor’s software fee rides on top of the infrastructure rate. Two parties are metering the same idle hours. The deployment behaves like the SageMaker endpoint it resembles under the hood: no traffic sensitivity, no stop verb, and hourly rates that, for large models on accelerated instances, make a forgotten deployment one of the fastest-accumulating mistakes in an AI account.

Third-party deployments under watch

ZopNight’s Bedrock provider discovers marketplace endpoints on the 6-hour cycle, with per-endpoint cost attributed from Cost Explorer or CUR 2.0. Idle-endpoint recommendations are the mainline finding: an endpoint whose invocations have flatlined is paying compute and vendor fees to hold a model in memory for nobody. The framing ZopNight adds is comparative. The same catalog model may be available on serverless terms elsewhere, or the workload’s actual volume may price far below the dedicated deployment, and per-endpoint cost data is what makes that arithmetic possible.

Evaluation leftovers and their invoices

Marketplace endpoints are disproportionately born from evaluations: a team trials a specialty model (a domain-specific LLM, a niche embedding model), deploys it for a bake-off, and the bake-off ends without a teardown. Vendor licensing sharpens the sting, since the software fee was priced for production value, not for a deployment whose last invocation was the benchmark suite. Multi-model comparisons multiply it by the number of contenders left running.

Deployment inventory

The Bedrock console’s Marketplace deployments view lists endpoints with model, instance configuration, and status. Endpoint invocation metrics settle idleness; for anything active, the honest question is whether volume justifies dedicated instances or belongs on per-token terms.

See it fire on your bill.

Connect an account read-only. The first findings land in minutes.

417 rule families across 353 resource types on 22 platforms. Every threshold, metric, and IAM action is documented on these pages before you grant anything.

417 rule families documented
353 resource types covered
read-only default access level
Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console· Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console·