Hugging Face
Hugging Face inference backends for chat, embeddings, speech and images.
Route requests to Hugging Face inference backends, such as hf-inference, fal-ai, nebius and together, through DeepintShield.
What works
- Chat, with every chat model also available through Responses
- Embeddings, speech (TTS) and transcription (ASR)
- Image generation on hf-inference, fal-ai, nebius and together (streaming on fal-ai only)
- Image edit on fal-ai
Limits to know: Request bodies are limited to 2 MB. fal-ai audio must be MP3. No image variation.
How to connect
Add your Hugging Face access token in Model Hub → AI Providers, create a Virtual Key, then call huggingface/<inference_provider>/<model_id>.
Read the Hugging Face provider docs. Requests get the same Virtual Key budgets, guardrails and logs as every other provider on the AI gateway.