Overview
Inference Providers is Hugging Face's routing layer to third-party inference companies. You call an OpenAI-compatible endpoint at router.huggingface.co with your Hugging Face token, pick a model from the Hub (and optionally a provider), and Hugging Face forwards the request to a partner such as a specialised GPU or LPU provider. More than 200 models are reachable this way.
Billing is pass-through: Hugging Face charges exactly the provider's rate with no markup and consolidates it on your Hugging Face bill. Every free account gets $0.10 of monthly credits and PRO users $2.00, and after that you buy credits to continue. Alternatively, set your own provider key; the provider then bills you directly and Hugging Face does not charge.
Organizations on Team or Enterprise plans receive $2 of credits per seat, can bill usage centrally with an X-HF-Bill-To header, set a spending limit and disable specific providers.
Pricing and plans
Free tier · no card
$0.10 of monthly credits for free users (subject to change); $2.00 for PRO users and per seat on Team or Enterprise.
At the limit Not stated The vendor does not say what happens at the limit, or we have not found it yet.
| Plan | Price | What you get |
|---|---|---|
| Free account | $0 | $0.10 monthly credits; pay-as-you-go after a credit purchase |
| PRO | $9/mo | $2.00 monthly compute credits usable across Hugging Face compute services |
| Team / Enterprise | $20/user/mo | $2.00 credits per seat, pooled; admins can set a spending limit |
| Routed requests | Provider rate | Billed by Hugging Face with no markup |
| Custom provider key | Provider bills you | Hugging Face does not charge; monthly credits do not apply |
What it does
- One token, many providers
- Route to partner providers with your Hugging Face token; no separate provider accounts needed.
- No markup
- Hugging Face passes through provider prices without extra fees.
- Bring your own key
- Use a provider key and be billed by the provider directly.
- Organization billing
- Bill requests to an organization or resource group; admins can set spending limits.
- OpenAI-compatible
- Works with the OpenAI SDKs and Hugging Face's own Python and JavaScript clients.
Strengths and limitations
Strengths
- Try many providers and models with one account and one bill.
- No markup over provider prices.
- Small monthly free credits for every account.
Limitations
- Free credits are tiny ($0.10 a month for free users).
- Provider choice and performance vary per model; prices are set by each provider.
- Monthly credits do not apply when using your own provider key.
- Spending limits are an organization feature (Team or Enterprise).
Who it suits
Good for
- Developers already using the Hugging Face Hub
- Comparing providers for the same open model
- Small experiments on open models
Look elsewhere if
- Production apps that need one provider's SLA and support contract
- Proprietary models from OpenAI, Anthropic or Google
Alternatives to Hugging Face Inference Providers
| Tool | From | Free option | Stages |
|---|---|---|---|
| Replicate Model hosting API (image, video, audio and LLMs) | $0.09/hr | ||
| GroqCloud Inference provider (open-weight models) | Usage-based | Free tier | |
| DeepSeek API AI model API (model maker) | Usage-based | ||
| OpenRouter Model router (unified API) | — | Free tier | |
| Mistral API (Mistral Studio) AI model API (model maker) | Usage-based | Free tier | |
| Scaleway Generative APIs Inference provider (EU-hosted open-weight models) | Usage-based |
Questions
Does Hugging Face mark up Inference Providers prices?
No. Its docs say Hugging Face charges the same rates as the provider, with no additional fees.
How many free credits do I get on Hugging Face Inference Providers?
Free users get $0.10 a month (subject to change); PRO users and Team or Enterprise seats get $2.00 a month.
Can I use my own provider API key?
Yes. Set a custom provider key and the provider bills you directly; Hugging Face does not charge for those calls.
Sources read for this page