DeepInfra Joins Hugging Face Hub as New Inference Provider, Expanding Serverless AI Capabilities

Hugging Face has significantly expanded its serverless AI inference capabilities with the integration of DeepInfra as a new Inference Provider on the Hugging Face Hub. DeepInfra, renowned for its serverless AI inference platform, introduces a highly competitive token — based pricing model and offers immediate access to a catalog of over 100 models, making it straightforward for developers to embed a wide array of AI functionalities into their applications with minimal setup. This initial integration specifically launches support for conversational and text-generation tasks on Hugging Face, enabling access to a suite of popular open-weight LLMs, including DeepSeek V4, Kimi — K2.6, and GLM-5.1.

The integration empowers users with flexible options for managing their inference requests directly through the Hugging Face Hub's website UI. Within user account settings, individuals can set their own API keys for registered providers, including DeepInfra, and arrange providers by preference. This preference ordering influences how compatible third — party inference providers are displayed on model pages, as well as the priority in code snippets.

For developers seeking programmatic access, DeepInfra's services are readily available through the established Hugging Face client SDKs. This includes `huggingface_hub` for Python (version 1.11.2 or newer) and `@huggingface/inference` for JavaScript. These SDKs simplify the process of interacting with DeepInfra — hosted models, allowing developers to authenticate using a Hugging Face token, which then automatically routes requests to DeepInfra. Furthermore, the robust integration extends to various Agent Harnesses, encompassing popular tools like Pi, OpenCode, Hermes Agents, and OpenClaw.

This strategic integration significantly reinforces Hugging Face's prominent role as a central hub for AI model discovery, development, and deployment. By incorporating DeepInfra's serverless platform, Hugging Face not only diversifies its service offerings but also intensifies the competitive landscape among serverless AI providers within its ecosystem. This heightened competition is anticipated to translate into more cost-effective solutions and enhanced service quality for developers across the board.

The billing structure is designed to offer clarity and flexibility to users. When developers choose to use a custom API key from an inference provider like DeepInfra, they are directly billed by that specific provider. Conversely, for requests routed through the Hugging Face Hub, users authenticate via their Hugging Face account and pay only the standard provider API rates, with no additional markup from Hugging Face itself. This transparent model ensures developers receive direct value without hidden costs. Moreover, Hugging Face's PRO users benefit from a significant advantage, receiving $2 worth of inference credits every month, which can be utilized across all supported inference providers.