Source: Amazon Bedrock — Provisioned Throughput
Bedrock's reserved-capacity option: hourly-billed model units with no-commitment, 1-month or 6-month terms. Custom models need it unless their base model supports on-demand deployment.
Contents
- What: vendor documentation
- Author: Amazon Web Services
- Date: undated, read 2026-11-14
- Original: Provisioned Throughput
- Related: Deploy a custom model for on-demand inference (read 2026-11-14)
Summary
Describes buying a fixed level of throughput for a base or custom model, measured in Model Units (MUs).
Key claims
- An MU guarantees a number of input tokens and output tokens per minute across all requests for one model.
- Billing is hourly per MU; price depends on model, MU count and commitment (none, 1 month, 6 months; longer is cheaper).
- Committed provisions cannot be deleted before the term ends; billing continues until deletion.
- The PT page says a customized model must be served through Provisioned Throughput. The related on-demand deployment page carves out an exception: models customized on or after July 16, 2025 from Amazon Nova Lite, Micro, Pro or 2 Lite, or from Meta Llama 3.3 70B Instruct, can use on-demand deployment (
CreateCustomModelDeployment) in specific Regions. - Per-MU throughput figures and limit increases go through the AWS account manager.