# Hugging Face Inference Providers (slim)

> Hugging Face's serverless router that sends model calls to partner inference providers such as Groq, Cerebras, Together and Fireworks through one Hugging Face token, with an OpenAI-compatible API at router.huggingface.co/v1 and Python and JavaScript clients.

- Full: https://www.anchorterminal.com/tools/hugging-face-inference-providers.md (~10,000 tokens) · this version ~1,730 tokens · JSON https://www.anchorterminal.com/tools/hugging-face-inference-providers.json · canonical https://www.anchorterminal.com/tools/hugging-face-inference-providers
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-07 · pricing https://www.anchorterminal.com/tools/hugging-face-inference-providers#pricing · reviews https://www.anchorterminal.com/tools/hugging-face-inference-providers#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**C · 54/100 · rank #920 of 1486 · #28 in Model APIs & inference · not agent-ready · confidence medium**


Assessment: One Hugging Face token reaches open models on 18 partner providers at the providers' own prices, with a public model list that shows each provider's price, context and tool support. Router rate limits, error responses and its own status component were not found in the reviewed documentation, and free accounts have had no included credits since 7 October 2026.

## Facts

- Kind: Model router · vendor: Hugging Face, Inc. · category: Model APIs & inference · legal entity: Hugging Face, Inc. · provenance 84/100
- Endpoint: `https://router.huggingface.co/v1` (HTTP)
- Auth: API key · pricing: Pay per use · x402: no · licence: Proprietary service under the Hugging Face Terms of Service, which name Inference Providers. The `huggingface_hub` Python client is Apache-2.0 and `@huggingface/inference` is MIT
- Benchmark probes: not run yet, so Performance is pending
- Live, from our pollers (not part of the score): up at 2026-11-07 18:20 UTC · uptime 30 days 100.0% (200 probes) · p50 108 ms · p95 196 ms over 24 hours
- Vendor status page: not read, no machine-readable status found (2026-11-07 04:47 UTC)
- Surface listed: The serverless router at router.huggingface.co/v1 (chat completions, a beta Responses API and task endpoints), the Hub API's `inference_provider` model filter and the Python and JavaScript clients, read from the docs, the router's model list and the clients' repositories on 6 November 2026. No account was made and no inference call was sent
- Not the same as Inference Endpoints: Inference Endpoints (listed as hugging-face-inference-endpoints in gpu-compute) deploys a model on dedicated instances billed by the hour. Inference Providers routes per-token calls to partner providers' shared infrastructure. They share a Hugging Face account and compute credits
- Providers: Baseten, Cerebras, Cohere, DeepInfra, fal, Featherless AI, Fireworks, Groq, HF Inference, Novita, Nscale, OVHcloud AI Endpoints, Public AI, Replicate, Scaleway, Together, WaveSpeedAI and Z.ai
- Provider selection: `:fastest` by throughput (default), `:cheapest` by output price, `:preferred` by the account's order, or a provider name. Team and Enterprise admins can disable providers and set a spending limit
- Custom provider keys: A provider key stored in Hugging Face settings is swapped in when routing, so the provider bills the call and included credits don't apply
- Data handling: Request body and response not stored when routed. Debug logs kept up to 30 days without user data or tokens. Each provider's own data policy applies
- Usage reporting: Usage by model and provider in account settings. An organisation usage API with daily requests, cost and token counts is Enterprise only
- Rate limits: None documented for the router. The Hub's own limits (1,000 API requests per 5 minutes for a free user, 2,500 for PRO) cover Hub APIs, resolvers and pages
- Prices: openai/gpt-oss-120b input, cheapest provider listed (DeepInfra) $0.037 per 1M tokens; PRO monthly compute credits $2 per month (plan)

- Scores: Reliability 25, Performance pending, Schema & documentation 62, Agent ergonomics 58, Security & auth 79, Payments & pricing 20, Task success pending, Maintenance & community 80, Transparency & trust 71 · total over the 7 assessed categories
- Why: Reliability, Hosted lines. · Schema & documentation, No OpenAPI document for the router (openapi.json paths answer 404). · Agent ergonomics, Router reading of the model checklist (tool use, structured output, caching, context, batch, SDKs, errors), as for openrouter. · Security & auth, Model reading. · Payments & pricing, No x402, MPP or L402 (0). · Maintenance & community, Model reading. · Transparency & trust, Closed service under Terms of Service that name Inference Providers, with an Apache-2.0 Python client and an MIT JavaScript client (20).
- Sources: 19, open questions: 5, both in the full twin
- Capabilities: inference.router, inference.open-weights, inference.llm
- JSON: https://www.anchorterminal.com/api/v1/tools/hugging-face-inference-providers.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/hugging-face-inference-providers.svg` or a link to https://www.anchorterminal.com/tools/hugging-face-inference-providers from the README at the root of github.com/huggingface/huggingface_hub, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.
- Support: Discord https://huggingface.co/join/discord; https://huggingface.co/support; https://huggingface.co/docs/inference-endpoints/support/faq; more in the full twin

## Before you call it

1. Read https://router.huggingface.co/v1/models before choosing a provider; it shows price, context and `supports_tools` per provider
2. Pin a provider with `model:<provider>` when tool calling or structured output matters, since support varies by provider
3. Use a fine-grained token with only the Inference Providers permission, and send `X-HF-Bill-To` to bill an organisation
4. Back off on 429 without assuming a Retry-After header, since none is documented for the router

## Connect

```bash
pip install huggingface_hub   # or: pip install openai, with base_url https://router.huggingface.co/v1
```

```bash
curl https://router.huggingface.co/v1/chat/completions \
  -H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-oss-120b:fastest","messages":[{"role":"user","content":"How many Gs in huggingface?"}]}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Tencent Cloud TokenHub | C | 59.5 | inference.router, inference.llm, inference.open-weights | https://www.anchorterminal.com/tools/tencent-tokenhub.min.md |
| Vertex AI (Gemini Enterprise Agent Platform) | AA | 88.7 | inference.llm, inference.open-weights | https://www.anchorterminal.com/tools/vertex-ai.min.md |
| Amazon Bedrock | A | 81.7 | inference.llm, inference.open-weights | https://www.anchorterminal.com/tools/amazon-bedrock.min.md |
| Azure OpenAI in Foundry Models | A | 80 | inference.llm, inference.open-weights | https://www.anchorterminal.com/tools/azure-openai.min.md |
| GroqCloud | BB | 74.5 | inference.llm, inference.open-weights | https://www.anchorterminal.com/tools/groq.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)