Free monthly limit reached

Inference Endpoints on the Hub
4

Is the provider fixed to Groq because the endpoint URL contains Groq?

You are not hitting a “model limit” in Colab. You are hitting a billing and credits limit in the Hugging Face Router path that calls Groq.

Your error shows the key detail:

  • URL: /static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fgroq%2Fopenai%2Fv1%2Fchat%2Fcompletions%3C%2Fcode%3E%3C%2Fli%3E
  • HTTP: 402 Payment Required
  • Message: “free monthly usage limit for groq … add pre-paid credits …”

That means the request is being routed through Hugging Face’s Inference Providers router and the provider being used is Groq (because the path literally includes /groq/). (Hugging Face)

Background: what “Inference Providers” really is

1) Colab Pro vs Inference Providers

  • Google Colab Pro buys you more or better Google compute (GPU/TPU time, RAM, runtime priority).
  • It does not buy you unlimited usage on external hosted inference APIs like Hugging Face Router or Groq.

So “I have Colab Pro” does not affect this 402.

2) Hugging Face Inference Providers uses “monthly credits”

Hugging Face runs a central “router” that can send your request to many providers (Groq, Together, Hyperbolic, Fireworks, etc.). (Hugging Face)

But usage is not unlimited. Hugging Face gives monthly dollar credits:

  • Free: $0.10/month (no pay-as-you-go)
  • PRO: $2.00/month (pay-as-you-go allowed)
  • Team/Enterprise: $2.00 per seat/month (pay-as-you-go allowed) (Hugging Face)

This is the most common reason people suddenly see 402 Payment Required after “it worked for months.” It simply means the free monthly credits are now exhausted. (Hugging Face)

3) Two billing modes exist and they matter

Hugging Face explicitly supports two ways to pay:

  1. Routed by Hugging Face

    • HF applies your monthly credits.
    • HF bills you (pay-as-you-go only if PRO/Enterprise). (Hugging Face)

  2. Custom Provider Key

    • You add your own Groq (or other provider) API key in HF settings.
    • HF credits do not apply. The provider bills you directly. (Hugging Face)

So a “free monthly limit reached” can be:

  • HF free credits exhausted (HF-routed), or
  • provider-side credits/quota exhausted (custom provider key), depending on how you configured it.

Why you see it “no matter which provider I choose”

Because your request is pinned to Groq by the URL:

router.huggingface.co/groq/openai/v1/...

If your code sets base_url to a Groq-specific router path, you will hit Groq every time, even if you think you selected another provider elsewhere. This is a very common configuration mistake. (Hugging Face Forums)

Correct “provider-switchable” usage uses the unified endpoint:

/static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fv1%3C%2Fcode%3E (Hugging Face)

Why creating a new token did not help

A Hugging Face access token is just authentication. It does not refill monthly credits. Credits are account-level and reset monthly per Hugging Face’s billing rules. (Hugging Face)

Also, there is a second class of failures where token permissions or missing billing setup triggers similar issues. Hugging Face staff explicitly mention that 402 can happen if there is no payment method on the account or the token lacks the right permissions. (Hugging Face Forums)

What to do next (simple decision tree)

Step 1: Confirm which “limit” you hit

  1. Check your Inference Providers Settings usage breakdown (past month by model and provider). Hugging Face documents that this view exists in settings. (Hugging Face)
  2. Check whether you set a custom Groq provider key in HF settings (Custom Provider Key mode). HF explains custom keys and that HF can “swap” auth when routing. (Hugging Face)

If HF-routed and you are on Free, the answer is usually: you used up the $0.10 monthly credits. (Hugging Face)

Step 2: Fix the “provider selection” problem in your code

If you want to switch providers, do not call /groq/... in the base URL.

Use:

from openai import OpenAI
import os

client = OpenAI(
    base_url="/static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fv1",
    api_key=os.environ["HF_TOKEN"],
)

This is exactly how Hugging Face’s “Getting Started” page shows the OpenAI-compatible setup. (Hugging Face)

Then select provider via the library’s provider parameter (if using InferenceClient) or via whatever provider selection mechanism your wrapper uses. Example from HF docs for InferenceClient explicitly sets provider="together". (Hugging Face)

Step 3: Choose how you want to pay

You have three practical options:

  1. Stay HF-routed, upgrade to PRO

    • You get $2 monthly credits.
    • You get pay-as-you-go after credits are exhausted. (Hugging Face)

  2. Stay HF-routed, but bill an Organization (if you have one)

    • HF supports org billing via X-HF-Bill-To or bill_to. (Hugging Face)

  3. Bypass HF Router and call Groq directly

    • Use Groq’s OpenAI-compatible API base URL: https://api.groq.com/openai/v1. (GroqCloud)
    • This avoids HF monthly credits entirely, and uses Groq’s own account limits and billing.

Groq’s official docs show the exact OpenAI client configuration for that base URL. (GroqCloud)

“Similar cases” online (what others report)

These are very close matches to your situation:

  • HF forum: users getting HfHubHTTPError: 402 and “exceeded monthly included credits,” including examples on router.huggingface.co/together/.... (Hugging Face Forums)
  • HF forum: same 402 but routed to another provider (hyperbolic) through the HF router, showing it is not just Groq. (Hugging Face Forums)
  • HF forum replies: staff/community note 402 can be tied to payment method missing or token permission settings. (Hugging Face Forums)

Reference links (copy/paste)

Hugging Face Inference Providers pricing and monthly credits:
https://huggingface.co/docs/inference-providers/en/pricing

Hugging Face “Getting Started” showing base_url=/static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fv1%3A
https://huggingface.co/inference/get-started

Hugging Face forum thread with 402 “monthly included credits” examples:
/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhugging-face-payment-error-402-youve-exceeded-monthly-quota%2F144968

Groq OpenAI compatibility and base_url=https://api.groq.com/openai/v1:
https://console.groq.com/docs/openai

Quick summary

  • Your URL contains /groq/, so you are using Groq even if you think you switched providers. (Hugging Face)
  • Hugging Face free tier has a monthly credit cap ($0.10) and Free has no pay-as-you-go. (Hugging Face)
  • New tokens do not reset credits. Credits are account-level. (Hugging Face)
  • Fix: use /static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fv1%3C%2Fcode%3E if you want switching, or call Groq directly at https://api.groq.com/openai/v1. (Hugging Face)