Pay per second. Not per token.

Simple pricing, faster models, less compute, real savings.

Free

$5 free credits

  • All models
  • Custom models
  • Full API Access
  • No credit card

Pro

Pay as you go

  • Starting at $0.001 / sec
  • No minimums
  • No idle costs
  • Customer support

Enterprise

Custom Pricing

  • Everything from pro
  • Volume discounts
  • SLAs & premium support
  • Custom optimizations & deployments

Start free. Scale to millions. No surprises.

Standard vs Premium

STANDARD$0.001/sec

Best for:

  • Image generation
  • Small-medium LLMs
  • Vision models
  • Most workloads

Hardware:

RTX 5090RTX 5080
PREMIUM$0.005/sec

Best for:

  • HD video generation
  • Large LLMs (70B+)
  • High-res diffusion
  • Multi-model pipelines

Hardware:

B200RTX PRO 6000

We route your job to the right tier automatically. Or you can also specify.

Why per-second beats alternatives

Transparent

You pay for compute, not arbitrary units. No "credits" that mean different things for different models.

Optimization = Savings

Our optimized models run faster. That means you use less GPU time. Automatically.

No Idle Costs

No hourly minimums. No paying for GPUs sitting unused. Just execution time.

Billing details

We track GPU-seconds in real-time. You're billed monthly, or when you hit a spending threshold. Add a payment method after you exceed the free tier.

Wall-clock time from when your job starts executing to when it completes. Queue time doesn't count. Setup/teardown doesn't count. Just execution.

Yes. Set a monthly cap in your dashboard. We'll alert you at 80% and stop jobs at 100%.

Our next-gen optimization makes models run faster. Faster = fewer seconds = lower cost. You benefit automatically.

Yes. Paste a HuggingFace URL or upload your custom weights. We optimize it, and the same per-GPU-second rates apply. No "custom model premium."

Ready to start building?

Join the waitlist. Launching soon.