BentoML Prediction Server

AI Platform & Data BentoML 1.4.39 Hardened AMI

BentoML 1.4.39 on Ubuntu 24.04 serves machine learning models as an HTTP prediction API on CPU. Package 0.1.3 includes a scikit-learn sample model so you can send a first prediction, then open the Swagger UI. The image does not call a model provider.

Get started View on AWS Marketplace

AMI release

0.1.3 (BentoML 1.4.39)

Recommended

t3.large

Storage

32 GiB+ encrypted gp3

Access

Web UI in your browser

Released

Oct 24, 2026

Delivery

Amazon Machine Image in your own AWS account

Serve machine learning models as an HTTP prediction API on CPU, in your own AWS account.

Bundled sample model

Send a first prediction within minutes, then explore the Swagger UI.

Token-protected HTTPS API with per-instance credentials. Access through Session Manager with no inbound ports.

Quick start

Full guide →

1. 1 Subscribe
2. 2 Launch the instance
3. 3 Run the walkthrough
4. 4 Open Swagger
5. 5 Call the prediction API

What is on the image

This guide matches Marketplace package 0.1.3 (BentoML 1.4.39). Check the listing for the version currently on offer.

• BentoML 1.4.39 on CPU, with a bundled scikit-learn sample model.
• Swagger UI on port 3000, started for the browser with sushi-bentoml-example.
• Prediction API on port 8448. A per-instance token protects it. sushi-bentoml-api reads the access file for you.
• A walkthrough, /opt/sushi/docs/demo.sh, and notes at /opt/sushi/docs/README.md.

Recommended instance

t3.large with at least 4 GiB of memory, at least 32 GiB of encrypted gp3 root storage, and IMDSv2. Leave inbound application ports closed. If you use SSH, limit it to an administrative address range you choose.

Billing

The listing states a software rate of USD 0.03 per instance-hour. Confirm the offer before you subscribe. EC2, EBS, and data transfer are billed separately. This image does not call a model provider.

Tutorial

Watch on YouTube.

Where to go next

• Getting started: launch, open Swagger, and send a prediction.
• Configuration: the two ports and the access file.
• Operations: backups and support.
• Troubleshooting: Swagger and the token.
• Software Sushi: the company site.

Upstream project

See the BentoML documentation. Software Sushi packages this image and does not claim upstream or AWS endorsement. Use /opt/sushi/docs/README.md for this build.

Last updated November 1, 2026 · Suggest an improvement