Skip to main content

AI Studio Billing

This guide provides a detailed overview of how billing works in Hyperstack AI Studio. It explains which services incur charges, how those charges are calculated, and where to find usage and cost data within the platform. It also introduces the two billing models used in AI Studio: token-based pricing for text-processing services and hourly pricing for tasks that rely on dedicated compute.

In this article​

Overview of AI Studio Billing Model​

Hyperstack AI Studio uses two billing models to accommodate different types of services:

  • Token-Based Pricing: Applies to services that process text, such as inference, Playground, model evaluations, and data synthesis. Billing is based on the number of tokens used:

    • Input Tokens and Output Tokens are priced separately.
    • Usage is calculated per 1 million tokens.

  • Hourly Pricing Based on Compute Used: Applies to compute-intensive tasks like fine-tuning. Billing is based on the duration of the job and compute used, charged at an hourly rate.

Billed AI Studio Services​

The table below summarizes each billable service in AI Studio, its purpose, and billing model:

ServiceDescriptionBilling Model
Fine-TuningTrains a model on custom datasets using dedicated compute resources.Hourly (Compute Used)
Data SynthesisGenerates synthetic datasets using prompt-based generation.Token-based
Serverless Inference & PlaygroundRun inference through UI Playground or API using base or fine-tuned models.Token-based
Model EvaluationsTest model performance using benchmark or custom criteria.Token-based

Accessing Billing Information​

To view your AI Studio billing data:

  1. Go to the Billing page in Hyperstack.
  2. On the Billing Overview page, view a summary of your current usage and total AI Studio costs.
  3. For service-specific details, navigate to the Resource Activity tab.
  4. Select any listed service Usage Report to view details (e.g., Fine-Tuning, Serverless Inference).

Within the usage reports you will see:

  • For Token-Based Services (Serverless Inference & Playground, Evaluation, and Synthesis):

    • Model – The model used during the request.
    • Cost Per Input Token – The charge applied per 1 million input tokens processed by the service.
    • Cost Per Output Token – The charge applied per 1 million output tokens processed by the service.
    • Total Tokens – Combined total of input and output tokens used.
    • Total Cost – Total cost for that usage entry.

  • For Hourly (Compute-Based) Services (Fine-Tuning):

    • Job Name – The display name of the training run.
    • Model Type – The base model used for fine-tuning.
    • Duration (Hours) – Elapsed training time.
    • Total Cost – Total cost based on usage time and compute rate.
    • Cost Per Hour – Hourly cost for the service.

Pricing and Account Balance​

  • Token usage is calculated every minute and billed after use.
  • Your AI Studio usage draws from a shared account balance used across Hyperstack services.
  • Access to services is suspended if your balance falls below the required minimum.
General Hyperstack Billing Policies

AI Studio services follow the same account, billing, and payment policies as the rest of Hyperstack. Learn more.


Back to top