Per-second billing — Hard budget caps — No minimums

Plans & Pricing

Choose your plan, load credits, use infrastructure. Per-second billing with hard budget caps.

Pro

$49/mo

+ compute usage via prepaid credits

  • All training methods (SFT, DPO, GRPO, LoRA, QLoRA)
  • Data pipeline (connectors, cache, datasets)
  • Model registry & inference deployment
  • 10 concurrent instances
  • 25 GPU concurrency
  • Email support
Most Popular

Team

$199/mo per seat

+ compute usage via prepaid credits

  • Everything in Pro
  • 50 concurrent instances
  • 100 GPU concurrency
  • Team workspace with RBAC
  • Custom domains for inference
  • Audit logs
  • Priority support (Slack)

Enterprise

Custom

+ compute usage via prepaid credits

  • Everything in Team
  • Unlimited concurrency
  • Dedicated GPU reservations
  • SLA guarantees (99.9%+ uptime)
  • SSO / SAML / HIPAA
  • Volume compute discounts
  • Dedicated support engineer

How Billing Works

Plan subscription is charged monthly and determines your concurrency limits, feature access, and support level. Compute usage (GPU time, storage, egress) is billed against your prepaid credit balance at per-second precision. Load credits in any amount ($50 minimum). When credits are depleted, charges go to your default card. Hard budget caps prevent surprise bills.

Compute Rates

Instance TypeVRAMRAMvCPUsInterconnectOn-DemandSpotsave 60%
RTX 4090Best value
24 GB64 GB16—$0.49/hr$0.19/hr
L40S
48 GB128 GB16—$0.89/hr$0.36/hr
A100 40GB
40 GB128 GB12NVLink 3$0.79/hr$0.32/hr
A100 80GB
80 GB256 GB16NVLink 3$1.10/hr$0.44/hr
A100 80GB ×4
320 GB1 TB64NVLink 3$4.32/hr$1.73/hr
H100 SXMPopular
80 GB256 GB26NVLink 4$1.49/hr$0.60/hr
H100 SXM ×4
320 GB1 TB104NVSwitch$5.88/hr$2.35/hr
H100 SXM ×8Full node
640 GB2 TB208NVSwitch$11.68/hr$4.67/hr
H200 SXM
141 GB480 GB32NVLink 4$2.49/hr$1.00/hr
H200 SXM ×4
564 GB1.9 TB128NVSwitch$9.88/hr$3.95/hr
H200 SXM ×8Full node
1.1 TB3.8 TB256NVSwitch$19.68/hr$7.87/hr

What are Spot Instances?

Spot instances use spare GPU capacity at a significant discount (typically 40-60% off on-demand rates). The tradeoff: if demand spikes, your instance can be reclaimed with 60 seconds notice. Best for training jobs with checkpointing enabled — if preempted, just resume from the last checkpoint. Not recommended for long-running inference servers that can't tolerate interruption.

Managed Storage

$0.05/GB/mo

Persistent volumes with auto-orphan protection. Ephemeral NVMe included with instance. BYO S3/R2/Wasabi also supported.

Egress

$0.00 first 1TB/mo

Then $0.01/GB. No surprise bandwidth bills. Includes SSH, Jupyter, and HTTP traffic.

Training

GPU cost only

No training surcharge. SFT, DPO, GRPO, LoRA, QLoRA — all included with your plan. Pay only for compute time.

Bring Your Own Provider

Already have an account with Vast.ai, Lambda, RunPod, Hyperbolic, or CoreWeave? Use your own API keys and benefit from your existing credits and reserved capacity. We handle orchestration, monitoring, zero-drop SSH, and the full dashboard experience — you just bring the compute.

FuturaNexus Managed

We provision GPUs from our network. Best availability, fastest launch, automatic failover. Prices as listed above.

BYO Provider

Use your own provider API key. Pay the provider directly for compute. We charge a flat 5% orchestration fee for dashboard, monitoring, and SSH proxy.

Ready to launch?

Choose your plan, load credits, launch GPUs in seconds.