Consistent 404 Not Found on Inference API for all models

Beginners
1

Hello,

I am consistently receiving a 404 Not Found error when trying to use the Inference API, and I believe there is an issue with my account’s permissions.

I have confirmed the following through extensive testing:

  1. My API token is correct and authenticates properly. A diagnostic script confirms this.

  2. The error occurs for all models I try, including fully public models like HuggingFaceH4/zephyr-7b-beta and gated models I have been granted access to, like mistralai/Mixtral-8x7B-Instruct-v0.1.

  3. The issue is not network-related. I have tried from different IP addresses and regions using a VPN, and the result is the same 404 Not Found error.

  4. The issue is not the request format. I have also tried adding a standard browser User-Agent header to my requests, and the error still persists.

It seems there is a specific issue with my account’s fundamental ability to access the Inference API service. Could you please investigate?

My Hugging Face username is: Shkila-coder

Thank you.

1 Like
2

Inference API has been revamped into Inference Providers, resulting in significant changes to its specifications and the models that can be used. The models currently deployed are these.

3

Thank you for the clarification, John6666.

Following your advice, I switched my code to use a model from the official “Inference Providers” list that you linked: HuggingFaceTB/SmolLM3-3B.

Unfortunately, I am still receiving the exact same 404 Not Found error, even with this confirmed-available model.

This seems to prove that the issue is not with model selection, but a fundamental problem with my account’s permissions to access the Inference API for any model at all.

1 Like
4

Hm. Endpoint URLs have also changed, so if you’re referencing old code, you may need to rewrite it. If you’re experiencing account issues, I recommend to contact Hugging Face via email first. website@huggingface.co

5

Hi @Shkila-coder Can you please check to make sure you’re authenticated? You’ll want to make sure you’re using afine-grained token with Make calls to Inference Providers permissions. More info here: Authentication.

1 Like
6

Hello, thank you for the specific instructions.

I have now done the following:

Deleted all old tokens.

Created a new fine-grained token.

Gave it the specific permission: “Make calls to Inference Providers” under User Permissions.

Unfortunately, I am still receiving the exact same 404 Not Found error when calling the API for a public model from the approved list (HuggingFaceTB/SmolLM3-3B).

1 Like
7

I can send you the code and my settings if needed.

1 Like