H100, A100, MI300X, and emerging accelerators. Pin your model to a hardware family.
Up to 1,024 GPUs in a single cluster. NVLink, InfiniBand, RoCE.
Replicate clusters across 47 regions. Sub-50ms anywhere.
Up to 400GbE between nodes. Tuned for distributed training and inference.
Add or remove nodes without downtime. Auto-scale to demand.
Your cluster. Your data. No noisy neighbors. No cross-tenant access.
Token-efficient, with volume discounts and burn incentives.
On-demand, single-tenant
On-demand, single-tenant
On-demand, single-tenant
H100, 50% discount
Pricing is illustrative. Final rates are governed by on-chain parameters and may vary based on network state.
Tier-based access. Higher stakes unlock better economics and more capacity.
| Tier | Required stake | Access |
|---|---|---|
| Startup | 150M QUBIC | Dedicated cluster, single region |
| Business | 500M QUBIC | Multi-region, custom hardware |
| Enterprise | Custom | On-prem, private infrastructure |
Drop-in compatible with the OpenAI SDK.
from aigarth import Aigarth
client = Aigarth(api_key="sk-aigarth-...")
# Provision a dedicated cluster
cluster = client.clusters.create(
name="heli-prod",
hardware="8x H100",
region="us-east-1",
replicas=3,
)
print(f"Cluster ready: {cluster.id}")
# Submit inference jobs
job = client.inference.submit(
cluster=cluster.id,
model="aigarth-reason-1",
prompt="Hello, world!",
)Everything in the standard tier, plus the things enterprises need.