On the Build plan, LLM requests fail with HTTP 429, category MaxGatewayCredits. Only one project is visible, and September inference usage totals approximately $0.54 at the rates displayed in Billing. Could you explain which usage exhausted the allowance and how we can verify the remaining credit balance?
What is the project id where you are seeting that error?
I’m seeing what appears to be the same issue on a separate Build-plan project.
Project ID: p_160htyzx0v5
My Billing dashboard also shows inference usage materially below the included allowance, but LiveKit Inference is returning:
inference_quota_exceeded
category: MaxGatewayCredits
remaining_limit: 0
max_allowed: 0
free_tier: true
STT is also returning HTTP 429 through the LiveKit inference gateway. I reproduced it with both:
deepgram/nova-3
and
assemblyai/universal-streaming
So this does not appear to be specific to one STT provider.
This looks very similar to Thomas’s reported case with approximately $0.54 of displayed usage. Could you also check whether the inference entitlement / MaxGatewayCredits state for my project has incorrectly been set to zero?
Same issue here on a Build-plan project, and I can rule out the account-wide credit explanation for my case.
Project ID: p_4xti3ntsw9l (EU region)
Every inference request from our agent fails with 429 from the gateway:
type: inference_quota_exceeded
error: LLM token credit quota exceeded
category: MaxGatewayCredits
remaining_limit: 0
current_usage: 0
max_allowed: 0
free_tier: true
status: QuotaStatusExceeded
project_id: p_4xti3ntsw9l
Session IDs: 1b306e3f-2cce-4237-b35d-4358e39279a1, 9ddd128f-b59f-4444-83e1-a951e5b02b8e, 438b518f-705d-4cfa-856b-7e36d798baf6, fedf6566-6c2c-414e-a81c-3b306d1a0c6f (2026-09-18, ~15:22 CEST).
Like @Antony_Esquivel in the STT thread, the key inconsistency is current_usage: 0 with remaining_limit: 0 — the gateway is applying a limit of 0, not reporting exhausted usage. STT over the gateway (wss://agent-gateway.livekit.cloud/v1/stt?model=deepgram/nova-3) returns 429 as well, so it is not provider- or model-specific.
The Billing page shows our project on Build with low peak usage (gemma-4-31b: 22% of 100 req/min, 12% of 600K tokens/min; nova-3 STT: 40% of 5 concurrent; gradium TTS: 60% of 5 concurrent).
We initially suspected the account-wide Build credit pool (as described in another thread), but we deleted the only other project under this account and the error persists, so that explanation is ruled out for us.
Could you check the inference entitlement / MaxGatewayCredits state for p_4xti3ntsw9l? From our side there is nothing left to fix — routing is correct (the error itself names the right project ID), and usage is well below the displayed limits.
I’m seeing the same issue on another Build-plan project.
Project: fitness-witness
Project ID: p_2qmgtyxsyy3
On September 18, 2026 around 10:17–10:18 AM Central (CDT), LiveKit Inference began failing with HTTP 429 across multiple categories.
One captured LLM failure:
type: inference_quota_exceeded
category: MaxGatewayCredits
quota_type: llm
current_usage: "0"
max_allowed: "0"
remaining_limit: "0"
status: QuotaStatusExceeded
project_id: p_2qmgtyxsyy3
session_id: 37b83bc8-994e-4019-bcf2-06ca9bd20d1d
STT was also failing repeatedly with HTTP 429 during the same period.
This is my only LiveKit project, and the dashboard shows monthly inference usage below the Build plan’s included allowance.
Because both LLM and STT are failing and the gateway reports max_allowed: "0", this looks like a project-level inference entitlement/quota issue rather than actual usage exhaustion or a model-specific limit.
Could someone from LiveKit please check whether the inference entitlement / MaxGatewayCredits state for project p_2qmgtyxsyy3 has incorrectly been set to zero?
I don’t have a captured request ID, but I do have the session correlation ID above and additional failing session IDs if useful.
Hmm, I am not aware of any issues. Let me review all these reports.
The team found the issue causing this. They are rolling out a fix now.
Hi Wilson, issue seems to persist:
14:29:00 WARNING livekit.agents failed to prewarm connection pool {"exception_type": "APIStatusError"}
14:29:00 WARNING livekit.agents LLM request rate limited by inference gateway {"model": "openai/gpt-4.1", "request_id": null}
14:29:00 WARNING livekit.agents failed to generate LLM completion: message="Error code: 429 - {'type': 'inference_quota_exceeded', 'error': 'LLM token credit quota exceeded, category: MaxGatewayCredits, remaining_limit: 0, current_usage: 0, session_id: \*\*\*\*, project_id: p_4xti3ntsw9l, status: QuotaStatusExceeded', 'hint': 'LLM token credit quota exhausted. Wait for the next billing cycle or upgrade your plan.', 'documentation_url': 'https://livekit.com/pricing', 'quota_type': 'llm', 'category': 'MaxGatewayCredits', 'current_usage': '0', 'remaining_limit': '0', 'max_allowed': '0', 'status': 'QuotaStatusExceeded', 'free_tier': 'true', 'session_id': '\*\*\*', 'project_id': 'p_4xti3ntsw9l'}", status_code=429, retryable=True, body=LLM token credit quota exceeded, category: MaxGatewayCredits, remaining_limit: 0, current_usage: 0, session_id: \*\*\*, project_id: p_4xti3ntsw9l, status: QuotaStatusExceeded, retrying in 2.0s {"llm": "livekit.agents.inference.llm.LLM", "attempt": 2}
✗ type='llm_error' timestamp=1790080140.7988667 label='livekit.agents.inference.llm.LLM' error=APIStatusError("Error code: 429 - {'type': 'inference_quota_exceeded', 'error': 'LLM token credit quota exceeded, category:
MaxGatewayCredits, remaining_limit: 0, current_usage: 0, session_id: \*\*\*, project_id: p_4xti3ntsw9l, status: QuotaStatusExceeded', 'hint': 'LLM token credit quota exhausted. Wait for the
next billing cycle or upgrade your plan.', 'documentation_url': 'https://livekit.com/pricing', 'quota_type': 'llm', 'category': 'MaxGatewayCredits', 'current_usage': '0', 'remaining_limit': '0', 'max_allowed': '0',
'status': 'QuotaStatusExceeded', 'free_tier': 'true', 'session_id': '\*\*\*', 'project_id': 'p_4xti3ntsw9l'}", status_code=429, request_id=None, body='LLM token credit quota exceeded,
category: MaxGatewayCredits, remaining_limit: 0, current_usage: 0, session_id: \*\*\*, project_id: p_4xti3ntsw9l, status: QuotaStatusExceeded', retryable=True) recoverable=True
14:29:00 WARNING livekit.agents failed to recognize speech: message='Invalid response status (429 Too Many Requests)', status_code=429, retryable=True, retrying in 2.0s {"stt": "livekit.agents.inference.stt.STT", "attempt": 1, "streamed": true}your agent
✗ type='stt_error' timestamp=1790080140.8230386 label='livekit.agents.inference.stt.STT' error=APIStatusError('Invalid response status (429 Too Many Requests)', status_code=429, request_id=None, body=None, retryable=True)
recoverable=True
What plan are you on and what is your project ID? Is it possible you are on the “Build” plan and have reached your free limit?
We need to do a better job at making this easier to understand. The place to look is in the the billing page. Unfortunately at the moment you have to copy the content off that page and ask an AI to calculate it for you. Since it is a free plan all the rows show $0.00 since you are not charged. That makes it hard to see what limits you have reached.
I am chatting with our team on how to make this easier to understand.
Okay, thank you to clarify this. I’ll switch my plan to the next tier. How many turns do I have with the ship plan?
Thanks — I pulled the September billing usage and calculated it using the rates shown on the page.
LLM input: ~$0.2993
LLM output: ~$0.0051
LLM cached tokens: ~$0.6084
TTS: ~$0.5461
STT: ~$0.1728
Total LiveKit Inference usage: approximately $1.6316.
The Build allowance is $2.50, so this appears to leave approximately $0.8684 remaining.
However, today the gateway is still returning:
MaxGatewayCredits
current_usage: 0
max_allowed: 0
remaining_limit: 0
Project: p_2qmgtyxsyy3
Can you check why the gateway is treating this project as having zero remaining entitlement when the billing usage appears to be below the Build inference allowance?
I will check with our billing team on it.
Are you getting an error?
Yes. The agent is getting HTTP 429 from LiveKit Inference:
type: inference_quota_exceeded
category: MaxGatewayCredits
quota_type: llm
current_usage: 0
max_allowed: 0
remaining_limit: 0
status: QuotaStatusExceeded
Project: p_2qmgtyxsyy3
We also saw repeated 429s from STT afterward. This is reproducible today using our previously working LiveKit Inference configuration.
I spoke with the billing team yesterday, identified an issue, and we’re fixing it. I will check in to see when that will be rolled out. Sorry about that.
Appreciate the assistance.
@Rich_O_Farrell I just checked with the team and the change was just applied so your account should now report accurately and work as expected.
Retested after the reported fix. The issue is still reproducible on project p_2qmgtyxsyy3.
All three LLM generations today returned HTTP 429:
inference_quota_exceeded
MaxGatewayCredits
quota_type=llm
current_usage=0
max_allowed=0
remaining_limit=0
Example post-fix diagnostic session ID:
471c677f-ecc4-42c2-ac3c-e3eb48f37dc0
This was after you confirmed the bug had been fixed. Could you check whether the fix has actually propagated to this project/account?