Databricks: Foundation Model API premium models blocked by workspace rate limit of 0

Question

Problem description

I am experiencing an issue where calls to the Foundation Model API premium hosted endpoints (Claude Fable 5 and Claude Sonnet 5) via the AI Playground are returning a 403 PERMISSION_DENIED error, stating that "The endpoint is temporarily disabled due to a Databricks-set rate limit of 0." The problem began on August 24, 2026, and is ongoing.

Environment

Azure Databricks Premium tier, serverless workspace in Canada Central; AI Playground using premium hosted Foundation Model API endpoints; no external storage involved.

What I've already tried

I have verified that the workspace is active, on the Premium tier, located in Canada Central, with the affected models listed and enabled. I also used Azure CLI to check workspace properties, confirming the SKU and location. Additionally, I checked the workspace settings and engagement with the Databricks support team to confirm that the quota and model availability are appropriate. However, the block persists, and the models are listed as disabled due to a rate limit of 0.

Current status

The issue appears to be caused by a Databricks-administered rate limit set to 0 for these endpoints, which prevents invocation. I am seeking assistance to understand why this limit is enforced and how to have it increased or removed to enable normal usage of the premium hosted Foundation Models.

Answer 1

Based on the information provided, this does not appear to be a workspace configuration, SKU, or serverless enablement issue. The key indicator is the error:

403 PERMISSION_DENIED: The endpoint is temporarily disabled due to a Databricks-set rate limit of 0

Because the message explicitly references a Databricks set rate limit, the request is being blocked before it reaches the model endpoint, even though the workspace is active, Premium tier, and the models are visible in the AI Playground.

Why this is likely happening?

A rate limit of 0 requests/minute typically indicates that access to the premium hosted model has been disabled through a Databricks-managed entitlement, quota, or regional availability control rather than through a customer-configurable workspace setting.

The fact that:

• Both Claude Fable 5 & Claude Sonnet 5 are affected
• The issue started suddenly on 24 Aug
• Workspace validation, SKU verification, and model visibility checks were successful

suggests a platform-level entitlement or quota change rather than a configuration problem within the Databricks workspace itself.

What to verify?

Review the following items for the affected workspace:

• Workspace ID
• Azure region (Canada Central)
• Foundation Model API entitlement status
• Premium model access for Claude Fable 5 and Claude Sonnet 5
• Effective quota/rate limit assigned to the workspace
• Any recent changes to regional availability or model access policies

Observations:

If the effective rate limit for the workspace is currently 0 RPM, no workspace level configuration change will restore access. The restriction must first be removed or updated at the entitlement/quota level before requests can be processed successfully.

Given that multiple premium models became unavailable simultaneously and are returning the same error, the evidence points more strongly toward an entitlement, quota, or regional-access issue than a problem with the Databricks workspace configuration itself.

---

Help make this community better for everyone: if this answer resolved your issue, please accept it or leave an upvote. If not, share more details in a comment so we can continue the discussion and find the right solution.