Repolode
Submit
beam-cloud

beam-cloud/beta9

Open-source Go/Python runtime for serverless AI workloads, offering GPU inference endpoints, sandboxes and background task queues with scale-to-zero.

Open-source alternative to:ModalReplicate

View on GitHubWebsite
Stars
1,826
Forks
171
Open issues
32
Last commit
5 hours ago

What is it?

Beta9 is an open-source runtime for deploying and scaling AI workloads without managing infrastructure. Developers write Python with decorators to define a model endpoint, a background task queue, or an isolated sandbox, and the runtime handles containers, scheduling and autoscaling.

The project focuses on fast container cold starts, using a custom container runtime, scheduler and embedded caching. Workloads can fan out to many containers in parallel, scale to zero when idle, and mount distributed storage volumes. It can run on Beam's cloud GPUs or on GPUs you bring yourself, and it can be self-hosted.

The repository is the engine behind Beam's managed cloud platform. Its Python client is published as beam-client.

Key features

  • 01Container cold starts in under a second via a custom runtime and scheduler
  • 02Fan out workloads to hundreds of containers in parallel
  • 03Autoscaling model endpoints with GPU options such as A10G and H100
  • 04Background task queues with retries, usable as a Celery alternative
  • 05Isolated sandboxes for running LLM-generated code
  • 06Scale-to-zero and distributed volume storage

Use cases

  • Serving custom ML models as autoscaling inference APIs
  • Running untrusted or LLM-generated code in isolated sandboxes
  • Replacing a Celery-style queue with resilient background jobs
  • Self-hosting a serverless GPU platform on your own hardware

Quick start

  1. 1

    Install the Python client

    pip install beam-client
  2. 2

    Create an account on Beam

  3. 3

    Follow the Getting Started Guide to onboard

Things to know

  • Licensed under AGPL-3.0, a copyleft license that requires sharing modified source code when offered as a network service.
  • Quickstart relies on the hosted Beam cloud account and onboarding; self-hosting setup is not described in the README.

Topics

#serverless#gpu#llm-inference#ai-infrastructure#autoscaling#sandbox#task-queue#python#autoscaler#cloudrun

FAQ

What is beta9?
Open-source Go/Python runtime for serverless AI workloads, offering GPU inference endpoints, sandboxes and background task queues with scale-to-zero. Beta9 is an open-source runtime for deploying and scaling AI workloads without managing infrastructure. Developers write Python with decorators to define a model endpoint, a background task queue, or an isolated sandbox, and the runtime handles containers, scheduling and autoscaling.
Is beta9 free to use? What license does it have?
beta9 is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
Is beta9 still maintained?
Actively maintained. The last commit to beta9 was on Oct 19, 2026.
What can I use beta9 for?
Serving custom ML models as autoscaling inference APIs Running untrusted or LLM-generated code in isolated sandboxes Replacing a Celery-style queue with resilient background jobs Self-hosting a serverless GPU platform on your own hardware
How do I install beta9?
Install the Python client: pip install beam-client Create an account on Beam Follow the Getting Started Guide to onboard
Where can I get beta9?
The source code and installation instructions are on GitHub: https://github.com/beam-cloud/beta9

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects