Choose a committed plan

Choose a committed plan for an endpoint that runs steadily for a long time. You commit to a fixed term and pay for it up front. An endpoint on a Committed plan costs the same whether it is busy or idle, so use an hourly plan for development.

Before you begin

• Permission to create model endpoints in the project. A Member needs a policy that grants create permission for the Inference service.
• Credit for the full committed price. A prepaid account needs it in full before the endpoint is created. For rates, see E2E’s pricing page.
• At least one worker, the minimum for a committed endpoint.

Choose the plan

1. On the create form, open Select Computing Resource and choose a machine and a size under Configuration.
2. Under Instance Pricing, choose a savings plan, such as 1 Month Savings Plan or 3 Month Savings Plan, instead of On-Demand. For the longer terms, select Show more pricing.
3. Set the workers. The committed workers are the minimum workers you set when you create the endpoint.
4. Select Deploy Endpoint. You confirm the terms in Committed Plan Confirmation before TIR creates anything.

Confirm the terms

The cost is the plan’s price times the committed workers. Tick the box to agree that:

• The cost is charged up front for the term.
• You cannot end the endpoint before the end date.
• If you stop early, you get no refund for the remaining days. The full amount is non-refundable once it is deducted.

Then choose what happens when the term ends, and select Proceed:

• Auto renewal of this committed plan starts a new term of the same plan.
• Auto start hourly billing for this machine moves the endpoint to the hourly rate of the same machine.
• Auto deletion of this committed machine deletes the endpoint.

What a committed endpoint does

• It runs for its whole term. You can stop only an hourly endpoint.
• It always keeps at least its committed workers, and cannot scale below them. Workers above that count run, and are billed, at the hourly rate of the same machine.
• Deleting it ends it at once, and the committed charge, already paid for the whole term, is not refunded.

Check the plan

Under Plan on the endpoint’s Overview, the Plan type is Committed, Committed Upto is the end date, and Renewal Preference is the end-of-term choice: Auto Renewal, Convert to Hourly Plan, or Auto Delete.

Change what happens at the end

Open Actions and select Update Inference, then Update Endpoint. The confirmation opens again with your saved choice selected. Choose another option and select Proceed. The update form leaves out the plan section for a committed endpoint, because you cannot change the plan there.

For problems with stopping or scaling a committed endpoint, see Troubleshooting Model Endpoints. For the questions people ask about billing, see the Model Endpoints FAQ.