# beam-cloud/beta9

Ultrafast serverless GPU inference, sandboxes, and background jobs

Repository: https://github.com/beam-cloud/beta9
Canonical: https://ross.abutalabs.com/products/beta9
Homepage: https://beam.cloud
Language: Go
License: AGPL-3.0
License Family: copyleft
Topics: gpu, ml-platform, cuda, fine-tuning, generative-ai, large-language-models, llm, distributed-computing, llm-inference, self-hosted, autoscaler, cloudrun, developer-productivity, faas, functions-as-a-service, paas, serverless, serverless-containers
Last push: 2026-08-19T19:00:48+00:00

## Health v2 (maintenance only)
Score: 89/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 98, release rhythm 86, longevity 73
- inputs: {"age_days": 1023, "days_push": 14, "days_rel": 14, "gap_med": 0, "n_releases_24m": 1130}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1755, forks 159 (observed 2026-08-28T04:05:32.085359+00:00)

## What it is
Beam (beta9) is an open-source serverless runtime for AI workloads, providing GPU inference endpoints, isolated sandboxes for running untrusted/LLM-generated code, and background task queues with sub-second cold starts. It offers a Pythonic SDK, a hosted cloud with on-demand GPUs, and the option to bring your own cloud (AWS, GCP, bare metal).

## Use cases
- deploy a serverless GPU inference endpoint for my model
- run LLM-generated code in isolated sandboxes
- run background jobs and task queues on GPUs
- autoscale ML inference with scale-to-zero
- self-host a serverless GPU platform
- run vLLM on an H100 without managing infrastructure
- fan out workloads across hundreds of containers
- run scheduled cron jobs for ML pipelines

## When to choose
- you want serverless GPU inference with sub-second cold starts
- you need secure sandboxes for executing untrusted or agent-generated code
- you want to deploy AI apps from Python code without writing YAML or Dockerfiles
- you need scale-to-zero billing for bursty ML workloads
- you want to run workloads across your own cloud accounts plus a managed cloud

## When to avoid
- you need a fully self-contained on-prem platform with no cloud dependency
- your workloads are long-running always-on services better suited to plain containers or Kubernetes
- you require a permissive license since Beam is AGPL-3.0
- you only need simple CPU-only function hosting without GPU support

## Facets
- artifact type: service
- maturity: active
- function: llm-inference, serverless, deployment, container-orchestration, gpu-computing, machine-learning, sdk, cli, scheduling, workflow-automation
- domain: cloud-computing, machine-learning, large-language-models, developer-tools, infrastructure-as-code, gpu-computing, self-hosted
- platform: python, go, cloud, self-hosted, cli
- tags: serverless-gpu, faas, sandboxes, cold-start, autoscaling, byoc, task-queues, scale-to-zero, agpl, docker

## Member repositories
- beam-cloud/beta9 (main) score 89

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:32.085359+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:27:54.532146+00:00, confidence not recorded.
  - readme: https://github.com/beam-cloud/beta9 (fetched 2026-08-28T04:05:32.085359+00:00, sha 53c82f5f89ce)
  - homepage: https://beam.cloud (fetched 2026-08-29T11:06:05.891629+00:00, sha f85558f7b5b2)
  - site_page: https://docs.beam.cloud (fetched 2026-08-29T11:06:05.894348+00:00, sha 8394447d8b3f)
  - site_page: https://www.beam.cloud/about (fetched 2026-08-29T11:06:05.897837+00:00, sha 4fc1562d0985)
  - site_page: https://www.beam.cloud/pricing (fetched 2026-08-29T11:06:05.896058+00:00, sha 6dfdd9eb444c)
- Data as of 2026-08-30T08:39:29.467469+00:00.