OVHcloud on Hugging Face Inference Providers

Overview

OVHcloud has been added as a supported inference provider on the Hugging Face Hub, allowing users to run models through OVHcloud AI Endpoints directly from model pages, the HF website, and the official Python and JavaScript SDKs.

How It Works

Website UI

In the HF account settings users can add their own OVHcloud API key for direct calls or leave the field empty to have requests routed through Hugging Face; they can also reorder providers by preference, which affects the widget and code snippets shown on model pages.

SDK Usage

The huggingface_hub Python package (version >= 1.1.5) and the @huggingface/inference JavaScript package accept a model identifier with the :ovhcloud suffix; authentication can be done with an HF token for HF‑routed calls or with an OVHcloud API key for direct provider calls.

Billing and Pricing

When using a personal OVHcloud API key, OVHcloud bills the user directly; when authenticating via Hugging Face, the user is charged the provider’s standard rates with no additional markup from HF. Hugging Face PRO subscribers receive $2 of inference credits each month that can be applied across providers.

Technical Capabilities

OVHcloud AI Endpoints provide access to frontier models such as gpt-oss, Qwen3, DeepSeek R1, and Llama through a fully managed, serverless API. The service runs on European data‑center infrastructure, ensuring data sovereignty and low latency for European users. It supports structured outputs, function calling, and multimodal processing for text and image inputs. First‑token latency is under 200 ms, and the platform handles both text generation and embedding workloads.

Feedback and Next Steps

Hugging Face invites comments on the OVHcloud integration via the dedicated discussion space at https://huggingface.co/spaces/huggingface/HuggingDiscussions/discussions/49.