Simple Pricing for GPU Credits

PlanPriceCreditsRate LimitIDEBest ForAction
FreeCURRENT$015/mo5 runs / 15 minNot includedTrying it outIncluded
Lite
$2/mo$1.60/mo
20% OFF
100/mo30 runs / 15 minIncludedOccasional experiments
Plus
$5/mo$4/mo
20% OFF
300/mo30 runs / 15 minIncludedStudents learning CUDA
Pro
$10/mo$7/mo
30% OFF
700/mo30 runs / 15 minIncludedDevelopers building projects

Have a promo code?

Sign in to redeem a promo code.

Frequently asked questions

What is a credit?

One credit pays for one execution on a real NVIDIA T4 GPU: a playground run, or one command in the IDE (run, build, PTX, nvidia-smi and so on). Compiling and running a program in the playground is a single credit.

Is cuda.live free?

Yes. Every account gets 15 free credits each month, enough to work through the tutorials and try the playground. Paid plans add more credits, a higher run rate limit and the IDE.

What is the IDE?

A lightweight IDE in the browser for projects with several files: an editor, error checking, runs on a real GPU, program input, data files, a PTX/SASS assembly view, profiling and benchmark sweeps. It is available to paid users on every paid plan. The playground is free for everyone.

Am I charged when my program times out or the GPU fails?

Infrastructure failures, such as a dropped GPU connection, are refunded automatically. Timeouts caused by your program are free the first two times in a row; every third consecutive timeout costs one credit. A normal run resets the count.

Do unused credits roll over?

No. Your balance resets to your plan's monthly allowance at the start of each billing period.

Can I change plans?

You can upgrade at any time. You pay only the price difference for the current period, and the extra credits are added immediately. Downgrades are not available; cancel and resubscribe instead.

What happens if I cancel?

Cancel any time from your dashboard. Your account returns to the Free plan and your balance is capped at the free allowance of 15 credits.

What GPU does my code run on?

Every plan runs on NVIDIA T4 GPUs (16 GB, Turing architecture) with CUDA 12, nvcc, PyTorch, CuPy and Triton preinstalled.

Sign in to upgrade your account and increase your limits.