Inference Providers docs

[Wauplin]

List taken from #1652 (comment).

TODO list:

• add link to announcement blog: https://huggingface.co/blog/inference-providers
• things to explain somewhere
  • "routed request" vs "direct calls"
  • mention it is production ready (wasn't the case with Inference API)
  • many providers: some are faster, some are cheaper, some are more reliable, some propose more models, some propose different tasks, etc.
• billing
  • case "routed request with HF token" vs "routed request with provider token" vs "direct call"
  • case "billing enabled" (Pay as you go) vs "not enabled" (only within free-tier)
  • no extra fee from HF
  • bill to organization
• website UI
  • search for models (UI): https://huggingface.co/models?inference_provider=fireworks-ai&sort=trending
  • search all models (UI): https://huggingface.co/models?inference_provider=all&sort=trending
  • set custom API key from provider (optional)
  • order providers by preference (apply to widget and code snippets)
  • widgets
  • playground hf.co/playground
• Pages: 1 provider == 1 doc page with 1 description + 1 list of supported tasks + 1 snippet
• add list of providers on landing page
• Page Hub API:
  • search for models (API): https://huggingface.co/api/models?inference_provider=fireworks-ai
  • get warm models (any provider): https://huggingface.co/api/models?inference=warm
  • get warm status for a model: https://huggingface.co/api/models/google/gemma-3-27b-it?expand[]=inference
  • get providers for a model: https://huggingface.co/api/models/Qwen/QwQ-32B?expand[]=inferenceProviderMapping
• Page: "how to be registered as a provider"? (Page: "how to be registered as a provider"? #1664)
  • get in touch!
  • JS client integration
  • Model mapping integration
  • Billing integration => how does it work
  • Python client integration
  • how to test your endpoints?
• tasks page with code snippets / API schema / recommended models / etc. (started in [Inference doc] Next gen inference snippets #1643)
  • add text to video (Add text-to-video page #1672)
  • split between "favorite tasks" (LLMs, T2I, T2V, etc.) and "other tasks"
• current pages:
  • index/intro => adapt
  • getting started => removed in favor of intro
  • Supported Models => removed
  • pricing and rate limits => merged with "billing"
  • security / data policy => kept with new wording
• rename https://huggingface.co/docs/api-inference vers https://huggingface.co/docs/inference-providers => selective Rename of api-inference => inference-providers #1666
• mention Inference Endpoints somewhere
• add nice visual of all providers Add inference providers banner #1669
• remove legacy headers Undocument x-wait-for-model and x-use-cache headers #1673

[404missinglink]

hey this link is 404 on the inference docs: https://huggingface.co/docs/inference-providers/hf-inference

this is from the quick links section on the "show code" when selecting hf inference.

[julien-c]

thanks @404missinglink, it should be https://huggingface.co/docs/inference-providers/providers/hf-inference, let me fix it

[julien-c]

@404missinglink fixed

[Wauplin]

Closing this issue as most items have been addressed!

[julien-c]

great job with this @Wauplin @hanouticelina ^