Skip to main content

Recharge and Billing

1. Recharge​

Currently, GPUGEEK supports online recharge or corporate bank transfer.

When choosing online recharge, enter or select the amount to recharge.

Click Next Step.
Scan the QR code with your mobile phone to complete the payment.

If you choose corporate bank transfer, you need to make the transfer according to the remittance information. Then click the [Upload Voucher] button, fill in the corresponding information, and upload the voucher. After the review and verification are approved, the funds will be credited.

2. Instance Billing​

Instance Billing: Instances on the GPUGEEK platform have three billing modes: Pay-as-you-go, Daily/Weekly/Monthly package, and Spot mode. The differences among the three are as follows:

  • Pay-as-you-go
    Pay-as-you-go charges based on the actual usage of the instance. This mode does not require upfront payment, but you need to ensure that the balance is greater than the usage cost of one hour for the instance (if there are running pay-as-you-go instances, the balance will deduct the costs of the running instances). When shutting down, the cost is settled by the second and deducted from the balance. After shutting down, the GPU is not reserved.

  • Daily/Weekly/Monthly package
    Daily/Weekly/Monthly packages are a prepaid billing mode for GPUGEEK cloud server instances. Fees for the selected period need to be paid upfront. This mode is suitable for stable long-term business needs, and the price is cheaper compared to pay-as-you-go.

    • Expiration Reminder: 7 days (for daily/weekly/monthly) before the service expires, the system will send a reminder notification.
    • Post-expiration Handling: Instances will be shut down immediately, and GPU resources released. In this mode, you can also choose auto-renewal, in which case the instance will not be shut down automatically.
    • Prepaid Time Conversion: Daily = 24h; Weekly = 7x24h; Monthly = 30x24h.
    • Changing Billing Mode: When changing the billing mode, the cost of the current usage time will be calculated and deducted based on pay-as-you-go, and the remaining prepaid amount will be refunded. Note that sometimes changing the billing mode may not be as cost-effective as the original mode.

  • Spot mode
    This mode also charges based on actual usage, but unlike standard pay-as-you-go, it allows starting instances at a price lower than the normal market rate. However, please note that instances may be shut down when resources are in short supply, so ensure tasks can handle such risks.

Each user has a limit on the number of instances that can be started.

3. Storage Billing​

Storage Pricing

3.1 Instance Storage Billing​

GPUGEEK charges for instance storage, calculated in RMB/hour·GB. If the instance is pay-as-you-go, storage fees are settled every hour, regardless of whether the instance is running. If not needed, the instance can be directly released.

Spot instances follow the same billing standard as pay-as-you-go.

If the instance is under a daily/weekly/monthly package, storage fees are included in the total fee and paid upfront. However, after shutting down, storage will continue to incur fees (charged hourly). If not needed, the instance can be directly released.

Note: Since storage fees continue after shutdown, please promptly release instances that are no longer needed.
Instance storage is charged based on the selected quota.

3.2 Cloud Disk Storage Billing​

GPUGEEK charges for cloud disk storage, calculated in RMB/hour·GB, deducted hourly from the account. Cloud disk storage is charged based on used capacity.

3.3 Image Storage Billing​

Backup images occupy storage capacity, and image storage is also charged in RMB/hour·GB, deducted hourly from the account. Image storage is charged based on used capacity.

4. Model Billing​

Public Models​

Public models are billed based on usage. Official models are billed according to the model task type, such as token count, number of generated images, or number of generated videos. When invoking personal public models, you are only charged for the actual runtime of each request, and not for model startup or idle time after startup.

For detailed pricing, please refer to Model Pricing.

Private Models​

Private models are only visible to you. When using or invoking APIs, billing is based on the online duration of the hardware configured for the model. Online billing duration includes startup time, runtime for processing requests, and idle time after startup. When the model no longer receives requests, it enters a cold start state after about 10 minutes, during which no hardware is occupied and no fees are incurred.

Model Deployment​

Model deployment is billed based on the number of running instances and their online duration. The number of instances scales automatically based on request volume. Online duration includes model startup time, idle time after startup, and actual request processing time.

Model deployment supports elastic scaling, allowing you to set the minimum and maximum number of instances. Scaling occurs approximately every 10 minutes based on the request volume in the past 10 minutes, adjusting the number of instances to meet demand.

5. Overdue Payment​

Please keep an eye on your account balance. If the balance is insufficient, we will notify you. If the balance cannot cover instance and storage fees, instances will be shut down. If your account is in arrears, we will retain your data for 7 days, after which it will be deleted. Please recharge promptly to settle debts and avoid impacting normal usage.