CPU Compute
Long-running CPU pods. AMD EPYC and ARM Graviton-class SKUs, per-second billed.
What it is
CPU Compute is for workloads that don't need GPUs: data pre-processing, model serving for small models, control planes, web back-ends, batch ETL. Same per-second billing as GPU Compute, same warm pool for fast boot, same checkpoint and auto-resume.
When to use it
- Pre/post-process data around a GPU training run.
- Run a model server for a CPU-feasible model (whisper-small, embedding models).
- Long-running back-end services for your AI product.
Quickstart
asc pods run --cpu 8 --memory 32Gi --image ghcr.io/me/api:latestclient.pods.run(image='ghcr.io/me/api:latest', cpu=8, memory='32Gi')await client.pods.run({ image: 'ghcr.io/me/api:latest', cpu: 8, memory: '32Gi' });Limits & quotas
| Limit | Default | Burst | Notes |
|---|---|---|---|
| Max vCPU per pod | 192 | — | Larger SKUs via bare metal |
| Max memory per pod | 1.5 TB | — | |
| Pod max lifetime | 30 days | — | Use Agents for indefinite |
Pricing
See pricing. Pay-as-you-go, billed monthly via Stripe.
API surface
POST /v1/pods— create (same endpoint, no GPU fields)
Required scope(s): pods:read, pods:write. See Scopes.
Security
All access is authenticated and scoped. See Auth & scopes and Network controls.