weightsapi.INFERENCEConsole
Navigation
Dedicated capacity

Rent a dedicated GPU for your model.

Choose from six GPU configurations, place a monthly rental order and pay with your account balance or crypto credit.

At a glance

GPU orders and checkout are open. Prepay one month; the service period starts when provisioning is confirmed. Availability and delivery dates still require confirmation.

Dedicated GPU rentals

Orders open

Choose a configuration, select your model and pay for one month through checkout.

Monthly prepaid rental. Capacity and model compatibility are confirmed before activation.

L40S

48 GB · GPU memory

$520 / per monthRent this GPU

A100

80 GB · GPU memory

$990 / per monthRent this GPU

H100 SXM

80 GB · GPU memory

$1,690 / per monthRent this GPU

H200

141 GB · GPU memory

$2,290 / per monthRent this GPU

2 × H100

160 GB · GPU memory

$3,210 / per monthRent this GPU

8 × H100

640 GB · GPU memory

$12,170 / per monthRent this GPU

Orders are saved immediately. Payment confirmation and GPU provisioning are separate steps; no machine is active until provisioning is confirmed.

Multi-GPU memory figures are totals across the GPUs, not a guarantee of a single GPU’s usable memory. Capacity depends on model precision, context length and concurrency. A quote and deployment check are required.

WeightsAPI rates

Choose a machine and place your order

Select a GPU configuration below, choose the model you want to deploy and review the monthly price at checkout. Your order is saved in your account so you can return to it later. The listed price covers one month of reserved inference capacity, without per-token charges for that machine. Capacity remains finite.

Match the configuration to your model

GPU memory must cover model weights, serving overhead and the attention cache. The operator confirms the model revision, precision, context window and concurrency before activation. Memory shown for multiple GPUs is their combined capacity; it is not the memory of one device. Selecting a model does not guarantee it fits every GPU configuration.

Pay with your account balance or crypto credit

Checkout uses the monthly USD price saved with your order. Pay explicitly from your available account balance after a confirmed top-up of at least USD 100. If credit is missing, use the crypto checkout and its fixed asset and network addresses, then return to the GPU order once the credit is confirmed. Each top-up has a USD 100 minimum.

A payment instruction or transfer awaiting verification is not confirmed credit. Keep the payment and order references, and return to the GPU checkout once the account shows sufficient available credit. Refreshing a payment cannot credit the account.

Follow the order until activation

Your order awaits payment, then provisioning. The GPU becomes active only after the operator confirms the deployment and supplies the endpoint. Your prepaid month starts on that activation date. Hardware availability and delivery timing require confirmation; placing or paying for an order does not guarantee immediate capacity.

There is no automatic renewal or recurring balance deduction. Before activation, you can cancel the order from checkout. A paid order is then refunded to your WeightsAPI account balance; cancellation does not send cryptocurrency to a wallet. Active service changes and end-of-period shutdown require operator handling.

Confirm the deployment scope

The operator must confirm hardware availability, region, model compatibility and delivery timing before activating a paid order. Shell access, backups, failover and automatic scaling are not included by assumption. Keep application credentials secure and check the selected model’s license. Review the service terms and support information for the current operating scope.

Need a next step?

Find the relevant guide or prepare the details of your issue.

Open support