# Hugging Face ## Kurzbeschreibung **“The AI community building the future.”** **Hugging Face **is not a single proprietary LLM provider, but a platform for hosting, discovering, distributing, evaluating, and deploying AI and LLM models. The Model Hub is used for storing, discovering, and using model checkpoints; LLMs can be used via Inference Providers, Inference Endpoints, or locally through libraries such as Transformers. ## Claim LLM “The AI community building the future.” ## Geeignet für - API Integration - Automation / Workflows - Education - Data Analysis - Data Extraction / Document Analysis - Email / Communication - Customer Service & Chatbots - Programming / Software Development - Research - Writing & Editing - Texts / Content - Translations - Science - Knowledge Management / Internal Search ## Kernfunktionen - Endpoints - EU storage - Function Calling - Inference - LLM API - MLOps - Model Router - Open-Source LLMs - PrivateLink - Provider Change - SSO - Structured Outputs ## Preismodell - **free:** You can test API access with a free Hugging Face account. There are monthly free credits. According to the current Hugging Face documentation, free users receive monthly credits, currently listed as $0.10, subject to change. After that, you need additional credits or pay based on usage. - **subscription:** **PRO **With Hugging Face PRO, you get significantly more included inference credits. The pricing page lists, among other things, 20× included inference credits for PRO; the Inference docs currently mention $2.00 in monthly credits for PRO users. **Team & Enterprise** For organizations, there are Team and Enterprise. These plans also include Inference Provider benefits or credits per seat and enable centralized billing, limits, and administration. According to Hugging Face, Team/Enterprise organizations currently receive $2.00 per seat in monthly credits. - **other:** **Pay-as-you-go** If your credits are used up, you can continue making API requests by purchasing additional credits or paying based on usage. The costs depend on the specific model, provider, and usage. **Your own provider key **In some cases, you can also use your own API keys from external providers. In that case, billing does not go through Hugging Face, but directly through the respective provider; according to the documentation, Hugging Face does not charge for this call. ## DSGVO und Datenschutz **Gesamteinschätzung:** Conditional **Overall assessment: **LLM router, API, and inference platform; not a traditional single proprietary LLM provider. As a pure LLM provider, Hugging Face primarily offers access to many models via Inference Providers, HF Inference, and Inference Endpoints. Inference Providers enable access to numerous external providers such as Cerebras, Cohere, DeepInfra, Fireworks, Groq, OVHcloud AI Endpoints, Replicate, SambaNova, Scaleway, Together, and others through a unified API. Access is integrated into SDKs for Python and JavaScript and, according to Hugging Face, can also be used via OpenAI-compatible API configurations. **Hosting model:** SaaS/API, serverless inference via Inference Providers, dedicated Inference Endpoints, protected or private endpoints, as well as EU/US storage regions for Team and Enterprise organizations. For Inference Endpoints, Hugging Face specifies three security levels: Public, Protected, and Private; Private Endpoints are accessible only via intra-regional AWS or Azure PrivateLink connections. **Data processing and training: **For Inference Providers, Hugging Face states that it does not store user data for training purposes and does not store request/response data for routed requests; logs are retained for up to 30 days for error analysis, without user data or tokens. For Inference Endpoints, Hugging Face states that it does not store payloads or tokens; logs are likewise stored for 30 days. However, external providers remain responsible for their own security and data processing. **Integrations:**Relevant here are Python/JS SDKs, Hugging Face InferenceClient, OpenAI-compatible API usage, Function Calling, Structured Outputs, and integrations into developer tools. This makes Hugging Face particularly strong as an LLM provider for applications where models need to be switched, compared, or connected across providers. **Conclusion: **As an LLM provider, Hugging Face is less a single model like Claude, Gemini, or GPT, and more an LLM infrastructure and routing platform. For developers and companies, this is powerful because a single API access point opens up many models and providers. For data protection and compliance, however, this means: not only Hugging Face, but also the specifically chosen Inference Provider must be reviewed. [Security & Compliance](https://huggingface.co/docs/inference-endpoints/security) **Overall assessment: **Conditionally GDPR-suitable. Considered purely as an LLM provider, Hugging Face is primarily GDPR-compliant when companies use Team or Enterprise features, a Data Processing Agreement, and a controlled infrastructure configuration. A positive aspect is that Hugging Face states for Inference Providers that it does not store user data for training purposes and, for routed requests, stores neither the request body nor the response; according to the documentation, debugging logs are retained for up to 30 days and contain no user data or tokens. In addition, TLS/SSL is provided for transmission, the Hub is SOC 2 Type II certified, and GDPR Data Processing Agreements are offered for Enterprise plans. **Negative **is that Hugging Face functions as a router to multiple external AI inference providers for Inference Providers; for the specific data processing, Hugging Face explicitly refers to the privacy and security policies of the respective provider. As a result, a blanket GDPR assessment for all LLM models and providers is not possible. The general Privacy Policy also names Hugging Face, Inc. and servers in the USA; personal data may be processed in the USA or other countries. Third-party providers or subprocessors listed include AWS, Google Cloud Platform, MongoDB Atlas, Stripe, GitHub, OVHcloud, and Hugging Face SAS. **Server location:** For the general service, Hugging Face lists servers in the USA; with Storage Regions, Team and Enterprise organizations can store repositories, models, datasets, and Inference Endpoints in EU data centers. However, for pure Inference Provider LLM calls, the actual processing location depends on the selected provider and its policies. **Conclusion:** For GDPR-critical LLM use, Hugging Face is not universally “simply safe,” but with an Enterprise DPA, EU storage, dedicated Inference Endpoints, PrivateLink, provider review, and clear logging/retention, it can be well controlled. For personal or confidential data, no arbitrary Inference Provider should be used without separate review. No verified information is available for uniform GDPR compliance across all connected LLM providers. [Security & Compliance](https://huggingface.co/docs/inference-endpoints/security) ## Hosting und Daten - **On-Prem / lokales Hosting:** abgedeckt - **Private Cloud / Rechenzentrum:** teilweise / indirekt - **EU SaaS / Managed:** teilweise / indirekt - **Hybrid:** abgedeckt - **AVV / DPA:** teilweise / indirekt - **Kein Training auf Kundendaten:** abgedeckt - **Open-Source / Transparenz-Pfad:** teilweise / indirekt ## Standort **Land:** France **Taxonomie:** France Hugging Face, Inc.: USA / Delaware Corporation; EU main establishment: Hugging Face SAS, 9 rue des Colonnes, 75002 Paris, France. ## Vorteile - Very large LLM/model catalog with community, research, and enterprise models - Unified API for many providers and model types - OpenAI-compatible entry point for chat completions - Dedicated Inference Endpoints for production deployments with autoscaling, logs, and metrics - Strong open-source libraries such as Transformers, Datasets, Tokenizers, PEFT, TGI, and Safetensors - Enterprise features such as SSO, RBAC, audit logs, resource groups, storage regions, and private repositories ## Nachteile - Not a classic “one-model-from-a-single-vendor” LLM provider; quality, licensing, and governance depend heavily on the respective model. - Community models and external providers require your own review of licensing, data protection, security, and model risks. - Inference Providers forward requests to external providers via a proxy layer; their data protection and security terms must be reviewed separately. - Pay-as-you-go and GPU-based usage can be difficult for beginners to estimate. - Scale-to-zero can cause cold starts and is therefore not suitable for all real-time applications. ## Quellen - Offizielle Website: https://huggingface.co/models ## Letzter Datenstand 2026-05-04 ## Originalseite https://kifox.ai/en/ki-tools/hugging-face-en/