Starter
$0 /mo
For prototypes and the first deploys to the mesh.
- 1 project, 2 regions
- Community models included
- $10 / mo compute credit
- Scale to zero, free
- Community support
Pricing
Pay for compute.
Pay for compute.
Not for complexity.
Usage-based pricing across the mesh. No region premiums, no egress traps, no per-seat tax. Scale to zero for free.
Most popular
Scale
$0.18 /GPU·hr
For teams shipping production AI to global users.
- Unlimited projects & regions
- Custom & private weights
- Agent runtime + data mesh
- $0.012 / GB·mo vector store
- 99.95% SLA, priority support
- Per-request observability
Enterprise
Custom
For regulated, high-volume, dedicated-fabric workloads.
- Dedicated GPU pools
- Bring your own cloud
- Region residency & DPA
- SOC 2 / HIPAA / BAA
- 99.99% SLA + TAM
- Architecture review & onboarding
Compute
$0.18 / GPU·hr
Per-second billing while warm. Free while scaled to zero.
Inference
$0.40 / M-tokens
Includes speculative decoding & KV-cache replication.
Data mesh
$0.012 / GB·mo
Vectors, objects, and lineage. Egress between regions free.
Compare
What you get on every plan.
Billing
Pricing questions.
No. Traffic between mesh regions is free — that's the point of a fabric. You only pay for compute, tokens, and stored data.
Compute billing stops. A small per-region keep-alive covers pre-warmed snapshots so cold starts stay under 800ms. On Starter, keep-alive is free.
Yes — commit to monthly GPU·hours on Scale or Enterprise for up to 32% off. Talk to sales for volume commits.
On Enterprise, workloads can run on your AWS/GCP accounts; you pay your cloud directly for that compute and Cyata for the control plane.
Try Cyata free for 14 days.
No credit card. Deploy your first model to 38 regions today.