Hugging Face Inference Endpoints vs Koyeb

Hugging Face Inference Endpoints

OAuth scopes separate reading endpoints from writing them, both OpenAPI documents are public, and the unauthenticated /v2/provider route lists every instance with its hourly price. An account needs a payment method and credits before the first deployment, no rate limits or SLA were found for the management API, and the docs price table disagrees with the live list in places.

Koyeb

H100 at $2.50 and H200 at $3.00 an hour, billed per second. Changelog silent since 27 February 2026 and no CLI release since 12 May 2026.

Hugging Face Inference Endpoints

1. Call GET https://api.endpoints.huggingface.cloud/v2/provider first and pick an instance whose status is available. The docs table lists types the API marks deprecated or not available
2. Send X-Scale-Up-Timeout: 600 on requests to an endpoint that scales to zero, or handle 503 while the first replica starts
3. Set scaleToZeroTimeout yourself. The docs give a default of 1 hour and the OpenAPI document says 15 minutes
4. Pause or delete an endpoint when the job is done. Billing covers every minute a replica is initialising or running
5. Give the agent a fine-grained token or the read-endpoints scope unless it must deploy. Endpoints are private by default and take the same Hugging Face token as a bearer

Koyeb

1. Pass dry_run on service create or update to validate the definition before anything deploys
2. Page service lists with limit and offset and filter by statuses instead of fetching everything
3. Expect the first request after deep sleep to take 1 to 5 seconds and retry once with a timeout
4. Set --min-scale 0 on GPU services so idle instances stop billing
5. Re-check the Mistral transition before building anything long-lived on it