Claude Fable quota stuck at 0 - every increase request auto-denied on a paid account with 20 months of usage

Google Cloud Build with AI Generative AI & Foundational Models
1

I’m hoping a Googler can trigger a manual review. Cloud Billing Support closed my case and referred me to the community forums, so this is the right venue.

Symptom

All Anthropic Claude Fable quotas on my project are 0 - requests/min and input/output tokens/min, across global, US multi-region and EU multi-region. Calls fail with 429 RESOURCE_EXHAUSTED - "Quota has been exhausted" on global_online_prediction_requests_per_base_model.

Already done

Model enabled in Model Garden, terms accepted, enablement questionnaire submitted. Publisher data sharing enabled via setPublisherModelConfig, confirmed dataSharingEnabledProvider: ANTHROPIC in both global and us-east5 - this cleared an earlier 403 that specifically demanded it. Model access is granted; zero quota is the only remaining blocker.

The problem

50 quota increase requests on this project, every one Denied, with response time equal to request time - same minute, so nothing human reviewed them. The most recent asked for 60 requests/min, 200,000 input TPM and 50,000 output TPM on global and US multi-region; all six line items were denied within a minute of filing.

Why the usual explanations do not fit

Not a fresh trial account: paid billing with continuous usage since January 2025 across 11 services, 2 projects, 66 SKUs.

Not rate limiting: the ceiling is 0 and current usage is 0%, so retry and backoff advice cannot apply.

Not a publisher-side hold: Anthropic support confirmed they do not set or gate per-project quota allocation for Claude on Vertex AI, and that no Anthropic-side hold exists on this project.

That leaves Google’s allocation process.

This is not isolated

See “Anthropic Claude quota stuck at 0 - requesting manual model-access review” (11 days old, no staff reply), where QuotaIncreaseEligibility returns NOT_ENOUGH_USAGE_HISTORY. Several older threads show the same pattern and none of them resolve.

Ask

Could staff trigger a manual quota eligibility review, or state what criteria the automated reviewer applies? If individual paid accounts cannot receive partner-model quota, I would rather hear that plainly than keep filing requests that are denied in under a minute. Happy to share the project number privately.

2

Update, with a concrete data point that narrows this down.

I queried Google’s own Cloud Quotas API for this project (`GET https://cloudquotas.googleapis.com/v1/projects/PROJECT/locations/global/services/aiplatform.googleapis.com/quotaInfos`).

Across all 367 aiplatform.googleapis.com quotas on this project, every single one returns quotaIncreaseEligibility.isEligible: true, with no ineligibilityReason set. Tally: 367 true, 0 false.

That includes the exact quotas whose increase requests are being denied. GlobalOnlinePredictionRequestsPerBaseModelPerMinutePerBaseModel, GlobalOnlinePredictionInputTokensPerMinutePerBaseModel, GlobalOnlinePredictionOutputTokensPerMinutePerBaseModel and UsMultiRegionResponsesRequestsPerMinutePerBaseModel all report isEligible: true.

So this project is not flagged NOT_ENOUGH_USAGE_HISTORY, or ineligible for any other published reason. Cloud Quotas reports it as eligible for increases on precisely the quotas whose increase requests are auto-denied within 60 seconds of filing, 50 times in a row.

That looks like an inconsistency between two Google systems rather than a policy decision: the eligibility service says the project is eligible, the approval pipeline denies without review, and neither surfaces a reason to the user.

Could staff look at why the quota increase pipeline rejects requests for a project that the Cloud Quotas API reports as eligible? Happy to provide the project number and the denied request IDs privately.

3

Hello,

I am experiencing the same Claude Fable 5.1 quota issue on my new Google Cloud project.

Project ID:
(PII Removed by Staff)

Fable 5.1 is enabled through Google Cloud Marketplace, the Anthropic terms are accepted, and Anthropic data sharing is enabled.

Authentication is working successfully, but API requests fail with:

429 RESOURCE_EXHAUSTED

Quota:
aiplatform.googleapis.com/us_multi_region_online_prediction_requests_per_base_model

Base model:
anthropic-claude-fable

My Cloud Quotas API shows the Fable base-model quota, but I am unable to obtain usable quota.

This is for low-volume personal research, development, and testing.

Could a Google staff member please review this project and advise whether an initial Fable 5.1 quota can be enabled?

Requested initial quota:
10 requests/minute

Thank you.