#BYOK 1 M free requests still showing 5 % deduction (Azure and Vertex key)

8 messages · Page 1 of 1 (latest)

muted needle
#

I’m on a personal account (no org) and added my Azure OpenAI adn Vertex Gemini BYOK. According to the announcement I should get 1 M free BYOK requests/month, then 5 % after that. However, every request I send is still being charged the 5% fee immediately and the $5 account balance is dropping.

Could you check if the free 1 M counter is enabled for Azure and Vertex keys, or if there’s an extra step I missed?

tidal lakeBOT
#

I'm an AI assistant - my answers may be incorrect and I cannot see screenshots. Please verify information with the OpenRouter docs.

I don't have specific documentation confirming that the "1M free BYOK requests/month" is live or enabled for Azure OpenAI and Vertex Gemini keys—only that BYOK is available and "is available for 5% of the upstream provider's cost" according to the official announcement. There are no additional setup steps mentioned; just add your keys in Settings.

Since your account balance is dropping from the start, I suggest:

  • Double-checking that your BYOK keys are active in your integrations.
  • Confirming if the 1M free tier is shown anywhere in your dashboard or usage pages.

If the free BYOK tier isn't working as described, this may be a billing or account-specific issue. Please email [email protected] for assistance—Discord does not handle billing issues.

Reference: "BYOK is available for 5% of the upstream provider's cost, while giving you all the other benefits of OpenRouter... Ready to get started? Head to Settings to add your keys." OpenRouter BYOK Announcement

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding.

The author you are looking for could not be found.

The model you are looking for could not be found.

muted needle
#

I checked the raw metadata of the requests and it said is_byok = false (using OpenRouter route), so I enabled Always use this key, and send request again and it had error "Error 400 (Azure) Provider returned error: Unsupported data type" now

tidal lakeBOT
#

Thanks for the update. Based on your description and the official documentation:

  • To ensure BYOK is used, you must enable “Always use this key” in your key settings. This forces OpenRouter to use your key for requests to that provider and prevents fallback to OpenRouter credits.
  • If you see is_byok = false in your request metadata, your key was not used and OpenRouter credits were charged instead.
  • The error "Error 400 (Azure) Provider returned error: Unsupported data type" suggests that the request sent to Azure is not formatted as Azure expects.

Regarding BYOK fees and free tiers:
The documentation states:
"The cost of using custom provider keys on OpenRouter is 5% of what the same model/provider would cost normally on OpenRouter and will be deducted from your OpenRouter credits."
There is no mention of a "1M free BYOK requests/month" in the official documentation, so I cannot confirm this feature exists for Azure or Vertex BYOK at this time. BYOK Documentation

For the 400 error:

  • Double-check your Azure key JSON configuration, especially the model_slug, endpoint_url, and model_id. Make sure your endpoint matches the format and API version required by Azure.
  • Review the exact data you are sending—Azure may require slightly different request payloads or data types compared to OpenAI.

If you continue to experience billing issues or configuration errors, please email [email protected] for further assistance, as billing/account issues are not handled on Discord.

Documentation references:

  • "Manage keys in your account settings."
  • "The cost of using custom provider keys ... is 5% ... and will be deducted from your OpenRout...

The model you are looking for could not be found.

#

...er credits."

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding.

west mountain
#

@muted needle have you been able to sort this oiut? I'm having the same problem

muted needle
#

Yup I figured it out

#

You should use response api instead of chat completions