Hugging Face Inference Endpoints

HF Inference Endpoints deploy models HF Hub managed.

Définition

Inference Endpoints supports text generation (LLMs), text classification, NER, summarization, translation, image classification, object detection, embeddings, ASR (Whisper), TTS. Auto-scaling 0-N replicas. Pay-per-second. Available clouds : AWS (us-east-1, eu-west-1), Azure (eastus, westus2), GCP. Service Distinct des Serverless Inference API et HF Inference Providers.

Origine

Hugging Face Inc., fondée 2016, Inference Endpoints lancé 2022, Series D 235M$ 2023 valorisation 4.5Mds$.

Usage

Standard pour déploiement managé modèles open-source HF Hub. Adopté Bloomberg, Pfizer, NASA, Mercado Libre.

Termes liés

• Hugging Face — plateforme parent.
• Hugging Face Hub — catalogue modèles.
• Together AI — concurrent.