Perpetual ML
Runs AutoML training, monitoring, and deployment in one platform that plugs directly into your data warehouse.
At a glance
Starts at
$0.90 per month
Price as published by the vendor, September 2026.
Free tier
No
Platforms
Web
Best for
Solo data scientists needing an end-to-end ML platform
Not for
Teams without Snowflake or a compatible data warehouse
3.8 out of 5
Scored by a Toolio reviewer after real useOur verdict
Perpetual ML bundles AutoML training with its own PerpetualBooster algorithm, experiment tracking, a model registry, drift monitoring, and both batch and real-time deployment into a single studio connected directly to your data warehouse. Continual learning claims to cut retraining time substantially as new data batches arrive, and a built-in Marimo notebook covers exploration alongside the automated pipeline. Pricing runs pay-as-you-go on compute, storage, and network egress rather than a flat fee, and native integration currently covers Snowflake, with Databricks still listed as upcoming.
✓What it does well
Automated model trainingPerpetualBooster automates training and claims leading AutoML benchmark performance.
Unified ML workflowExperiment tracking, a model registry, monitoring, and deployment all live in one platform.
Data warehouse integrationIt connects natively to Snowflake, working directly with data you already have.
✕Where it falls short
Usage-based cost adds upPricing is metered by compute, storage, and egress, so costs are not fixed the way a flat subscription fee would be.
Databricks not yet supportedNative integration currently covers Snowflake only, with Databricks support still listed as upcoming.
No free tierThere's no free plan, so usage-based charges start from the first training run.
Key features
AutoML trainingPerpetualBooster automatically trains models from your data.
Experiment trackingIt tracks, compares, and reproduces every training experiment in one place.
Drift monitoringIt monitors data and model drift without needing retraining or ground truth.
Batch & real-time deploymentTrained models deploy for either batch or real-time inference from the same platform.