Pricing

You get full access to the Bento Inference Platform—deploy open-source LLMs, or your own models with the BentoML open-source serving library. We’ll credit your account with one-time free compute credit so you can spin up GPUs and test deployments. Upgrading to the Starter plan (with a credit card) unlocks additional GPU types and lets you create more deployments.

When you create a deployment in Bento’s web console, you’ll see an hourly rate estimate. If you enable autoscaling, that estimate becomes a range based on your workload—and you have full control over autoscaling behavior (min/max replicas, scaling policies, queuing, replica concurrency, etc.). We meter by the second and only charge for active compute—deployments scaled to zero incur no cost.

No--deployments only incur charges while they’re running. Once they scale to zero, billing stops automatically.

We offer formal SLAs on the Scale and Enterprise plans, including cold-start guarantees and uptime targets. Contact us to learn more.

Starter customers get the North America region by default. Contact us if you need other regions or multi-region support.

Today, we don’t charge extra for network egress or storage—everything’s included in your compute rate.

Bring-Your-Own-Cloud (AWS/GCP/Azure): Provisioning usually takes just a few hours, depending on your security review, network setup, and quota availability. On-Premises: Full deployment typically completes in 1–2 weeks, depending on your infrastructure.

Starter: We charge your credit card monthly for actual compute usage (per-second metering). You’ll receive an invoice at month’s end. You can cancel anytime—just keep a valid card on file. Scale & Enterprise: We offer custom invoicing, payment terms, and consolidated billing. Contact sales to discuss the options that fit your organization.

All your usage metrics and detailed cost reports can be found in your account's billing page (https://cloud.bentoml.com/billing).

Starter: Email support and access to our community Slack. Scale: Everything in Starter plus a dedicated Slack channel and faster response times. Enterprise: 24/7 support with Zoom office hours, priority SLAs, and dedicated engineering resources.