The short answer
A charge from BASETEN* ML DEPLOY corresponds to serverless machine learning model deployment, Truss framework packaging, or dedicated GPU inference from Baseten, Inc.
Why it shows up like this
A charge labeled BASETEN* ML DEPLOY on your credit card or banking statement comes from Baseten, Inc. for Baseten Production ML Model Serving & GPU Hosting. Baseten is a cloud infrastructure platform that enables engineering teams to serve machine learning models on dedicated GPU hardware. On billing statements, the transaction commonly appears as BASETEN* ML DEPLOY, BASETEN INC, BASETEN SAN FRANCISCO, or BASETEN.CO.
Billing on Baseten is usage-based and reflects actual compute consumption. The platform bills per millisecond of model execution or per hour for reserved GPU instances such as NVIDIA A10G or A100 graphics cards. An unexpected invoice often occurs when a model packaged with the Truss framework is deployed to production, an API endpoint receives high web traffic, or a developer leaves a dedicated instance active without enabling scale-to-zero settings.
To locate the source of the charge, log in to the web console at app.baseten.co. Navigate to Settings and click Billing to inspect your consumption metrics and itemized monthly invoices. If you want to stop ongoing charges immediately, open the Models tab in your dashboard, select each active model, and click Deactivate. Deactivating models removes the provisioned instances and stops future compute billing.
Baseten considers consumed GPU compute time and inference requests non-refundable because cloud hardware was already provisioned. If a developer on your team set up the deployment with a company card, you can set spending limits inside workspace settings to prevent unexpected charges. If nobody in your company uses Baseten and you suspect card fraud, contact your card issuer to dispute the unauthorized transaction and obtain a new card.
Don’t recognize it? Common scenarios
- An AI model packaged with Truss was deployed to Baseten's production inference cluster.
- An autoscaling model endpoint received continuous traffic from an external web app.
- A dedicated GPU instance remained active on the account.
How to locate the invoice and cancel
- Log in to Baseten Dashboard Open app.baseten.co and sign in with your email or GitHub credentials.
- Check Workspace Billing review to Settings > Billing to review inference hours, GPU types, and download past invoices.
- Deactivate or Delete Models Go to Models, select your deployed model, and click Deactivate to scale instances to zero.
Frequently Asked Questions
How does Baseten handle scale-to-zero?
Baseten models configured with autoscaling scale down to zero active instances when idle, incurring no compute costs during inactivity.
Are Baseten GPU inference charges refundable?
Usage fees and reserved GPU hours are non-refundable once consumed. Deactivating models stops future billing.
Independent reference — not affiliated with Baseten, Inc.. Billing names and policies change; verify with the merchant or your bank before acting.
Verified sources
Every claim on this page is checked against official sources — open them to confirm before you call your bank.
- Baseten Pricing & Dedicated GPU Rates https://www.baseten.co
- CFPB: Disputing Unrecognized Charges https://www.consumerfinance.gov
Reviewed Aug 15, 2026 · high · About UnknownCharges