Simple, transparent pricing
BYO GPU is free — we manage the platform, you bring the hardware. Need more compute? Pay as you go, billed per-second in AUD. No reserved instances, no lock-in.
BYO GPU
Free
Bring your own hardware. Share compute across users, teams, and institutions — we handle the platform.
- Connect your GPUs to the Yarn network
- Share compute between teams and researchers
- Institutional SSO — researchers sign in with their university
- Manage compute distribution across an entire institution
- Full inference and training API access
- Overflow to managed pool when you need more
Managed compute
From $1.50/hr
Set clear budgets and plan your spending. Per-second billing, full cost visibility, no surprises.
- RTX 4090 and H100 GPU access
- Per-second billing — no idle charges
- Budget limits and alerts at user, team, and institution level
- Training jobs, sessions, and notebooks
- Chat and API inference
- Auto-reload and spend controls
GPU compute rates
Credits are consumed per-second of GPU time. No minimum usage, no idle charges.
| GPU | VRAM | Sovereign | Overseas |
|---|---|---|---|
| RTX 4090 | 24 GB | $1.50/hr | $1.20/hr |
| A100 40GB | 40 GB | $3.50/hr | $2.80/hr |
| A100 80GB | 80 GB | $5.00/hr | $4.00/hr |
| H100 SXM | 80 GB | $8.00/hr | $6.50/hr |
Rates are in AUD. Sovereign compute runs on bare-metal infrastructure in Perth, Western Australia. Overseas rates are indicative — actual pricing depends on provider availability.
One platform, multiple GPU providers
Yarn manages your compute across sovereign Australian hardware and global GPU cloud providers. Same API, same SDK, same budget controls — regardless of where the GPU lives.
Sovereign jobs stay in Australia. When you need more capacity, Yarn provisions cloud GPUs automatically and routes your workload — you never manage a cloud account.