Beam
Cloud that sources bare-metal servers and sells them as persistent VMs, sandboxes for untrusted code, task queues, serverless functions and endpoints, with GPUs from RTX 4090 to B200, billed per second while code runs; prices from $0.06 an hour and H100s from $1.83 an hour are published.
What it does
- GPU cloud
- Model hosting
- Agent orchestration
Features 1
- SandboxesIsolated computers for untrusted code that start in about half a second and can be paused, snapshotted and restored.
Sources
- Pricing
-
The pricing page lists plan cards "Developer $20/month $20 in credits each month" and "Team $89 Per month, plus usage" beside the per-second rates and a Growth plan on "Contact Us"; the recorded models omit subscription. Read 2026-11-16 with curl.
- Description
-
Official documentation · 9 Nov 2026
Product page: "Compute for agents"
- What it does
-
Official documentation · 9 Nov 2026
Product page: "Compute for agents"
- Sold within
-
Sold on its own with published plans.
- Name
-
Official documentation · 9 Nov 2026
Name as published on the product page.
- URL
-
Official documentation · 9 Nov 2026
Product page, read 2026-11-09.
- Deployment
-
Official documentation · 9 Nov 2026
Home page: a hosted service with an install script and SDK, and a BYOC option on the pricing page.
- Interfaces
-
Official documentation · 9 Nov 2026
Home page: a hosted service with an install script and SDK, and a BYOC option on the pricing page.
Also from Beam
-
Beam Sandboxes Infrastructure service
Isolated cloud sandboxes for agent- and LLM-generated code that start in about half a second, with memory snapshots, GPU attach and port exposure, billed by the millisecond.
-
Beam Virtual Machines Infrastructure service
Persistent Linux microVMs for AI agents, each with its own kernel, durable disk, SSH and browser desktop, that can be paused with memory intact and forked, billed per second while running.
-
Beam Serverless GPU Infrastructure service
Serverless GPU endpoints that turn a Python function into a model-serving endpoint on RTX 4090 and RTX 5090 GPUs that start when a request arrives and stop after traffic stops, billed by the millisecond.
Something wrong here? Send a correction — quote this product id: beam-cloud.