SaaS
Amazon Web Services
SageMaker Model Monitor stops accepting new customers and enters maintenance-only mode — no new features
departs 30 Jul 2026
What this means for you
This is a soft sunset, not a shutdown, and the distinction matters. From 30 July 2026 you can no longer start using SageMaker Model Monitor if you are not already a customer — new customer access closes. If you already run Model Monitor (data-quality, model-quality, bias-drift or feature-attribution drift monitoring schedules), AWS says you "can continue to use the service as normal," so nothing breaks on the date. What ends is the roadmap: AWS "do not plan to introduce new features," only security and availability fixes. So the real deadline is strategic, not operational — any team standing up new drift monitoring on SageMaker should stop and pick the replacement instead of building on a service that is now closed to newcomers and frozen.
Where to go
AWS names the replacement directly: a combination of its open-source SageMaker AI monitoring solutions (published in the aws-samples GitHub organization, built on Evidently AI and SageMaker MLflow), Amazon QuickSight governance dashboards, and Amazon CloudWatch. The open-source solutions cover the same ground — data-quality and model-quality drift, bias drift, feature-attribution drift — plus LLM evaluation and GPU observability, and run inside your own AWS account. The docs include step-by-step guidance for deleting existing Model Monitor schedules (which stops their recurring Processing-Job compute cost) and standing up the Evidently-based batch or real-time pipelines in their place.
- Open-source Amazon SageMaker AI monitoring solutions (aws-samples, Evidently AI + MLflow) — the vendor-recommended replacement
- Amazon QuickSight governance dashboards + Amazon CloudWatch anomaly detection
Amazon SageMaker Model Monitor watches machine-learning models in production for drift — it compares live inference data and predictions against a training baseline and alerts when data quality, model accuracy, bias, or feature attribution moves outside set constraints. AWS has now decided to wind it down for new adoption. Its documentation carries a note, repeated across the Model Monitor pages: “After careful consideration, we have made the decision to close new customer access to Amazon Sagemaker Model Monitor, effective 7/30/26.”
The wording is deliberately narrow, and reading it precisely is the whole point. This is not a service being switched off. Existing customers “can continue to use the service as normal,” and AWS says it “continues to invest in security and availability improvements.” Two things end on 30 July: the ability for a new customer to start using Model Monitor at all, and any future feature work — AWS states it does “not plan to introduce new features.” A frozen, closed service is one you can keep running but should not keep building on.
The good case here is that the replacement is named and documented rather than left as an exercise. AWS points to a combination of its open-source SageMaker AI monitoring solutions — built on Evidently AI and SageMaker MLflow, published under the aws-samples GitHub organization and running entirely in your own account — together with Amazon QuickSight dashboards for governance and Amazon CloudWatch for system- and inference-level metrics. The migration docs walk through deleting your existing Model Monitor schedules (which also stops their recurring Processing-Job cost) and rebuilding the equivalent data-quality, model-quality, and bias-drift checks on the Evidently-based pipelines. If you were about to adopt Model Monitor, the honest move is to skip it and go straight to the open-source path; if you already depend on it, you have time, but the roadmap has stopped moving under you.
Entry changelog
- — last reviewed.
- — Entry created from the official AWS SageMaker Model Monitor availability-change notice.
Spot an error or a changed date? Tell us and it’s corrected the same night.