According to the Hugging Face announcement, DeepInfra is now a supported Inference Provider on the Hugging Face Hub, expanding the platform’s serverless inference ecosystem. The integration enables developers to run a catalog of over 100 models directly from model pages, with support for large language models, text‑to‑image, text‑to‑video, embeddings and other capabilities. DeepInfra’s pricing is described as one of the most cost‑effective per token in the industry, and the service is accessible through the huggingface_hub Python SDK (version 1.11.2 or higher) and the @huggingface/inference JavaScript package. Users can authenticate with a Hugging Face token to route requests automatically, and PRO subscribers receive $2 worth of inference credits each month, while free signed‑in users receive a limited free quota. The provider also integrates with popular Agent Harnesses such as Pi, OpenCode, Hermes Agents and OpenClaw, allowing models hosted on DeepInfra to be used without extra glue code. Billing is handled by the provider for direct requests, and routed requests incur only standard provider API rates without markup. The announcement highlights upcoming support for additional tasks and invites feedback through the Hugging Discussions space.
DeepInfra on Hugging Face Inference Providers 🔥
DeepInfra is now a supported Inference Provider on the Hugging Face Hub, expanding serverless inference options with over 100 models and cost‑effective pricing. The integration works with Hugging Face SDKs for Python and JavaScript and is available to PRO users with monthly credits. Additional model types and tasks are planned for rollout.
Why it matters
This addition increases model accessibility and reduces cost for developers using Hugging Face, fostering broader adoption of AI services. It also strengthens the ecosystem of serverless inference providers, encouraging competition and innovation.
Topic dossiers Hugging Face