Hugging Face Pricing Update November 2022

Hugging Face has shifted its monetization strategy to focus on providing direct access to compute for AI through services like AutoTrain, Spaces, and Inference Endpoints. This change ensures that while basic inference remains free, users requiring enterprise-grade performance and custom hardware configurations can access scalable, paid compute resources directly from the Hub.

Transition from Inference API Paid Tier to Inference Endpoints

Inference Endpoints provide a fast, enterprise-grade inference-as-a-service solution, replacing the previously available Paid tier of the Inference API. While the standard Inference API remains available for all users for free, Inference Endpoints are designed for those requiring higher performance and stability for production environments.

Hardware Upgrades for Hugging Face Spaces

Users can now select specific hardware configurations to run machine learning demos on Hugging Face Spaces. This allows developers to optimize the performance of their ML demos based on the hardware requirements of their specific models.

Billing and Account Management

All paid services, including PRO subscriptions, hardware upgrades for Spaces, and Inference Endpoints, are managed through a centralized billing settings page.

• Payment Method: No subscription is required to use compute services; users only need to add a credit card to their account or attach a payment method to an organization.
• Invoicing: Usage for all paid services and subscriptions is charged at the start of each month, with a consolidated invoice provided for record-keeping.
• Usage Tracking: The billing settings page allows users to visualize their usage for the past three months.

Summary of Paid Compute Services

Service	Purpose
Inference Endpoints	Enterprise-grade, fast inference as a service
Spaces Hardware Upgrades	Custom hardware for ML demos
AutoTrain	Simplified access to compute for AI training