Hey @mikelben !
According to an official resolution provided by a Databricks Employee on the community forums, this specific "Permission_Deneid" error occurs when you are in the free/trial phase.
The "rate limit of 0" is a backend restriction that typically triggers if your workspace is still flagged under a trial or free (often placed as TRIAL_VERIFIED), which limits high-demand premium models (like GPT-5.x and premium Claude variants). Even with a valid credit card on the account, the workspace itself may not have been fully converted yet.
To get this resolved before you consider changing your architecture, you have two direct paths forward:
- Contact Databricks Support / Sales: Reach out to them explicitly asking to have your AWS workspace converted and fully verified for premium hosted foundation models (moving it to the PAYABLE_VERIFIED tier). Once they lift the workspace-level restriction, the endpoint should begin working immediately without any configuration changes.
- Use Un-gated / Open-Source Models (Workaround): In the meantime, you can test your architecture using open-source models like Llama, which are not subject to this specific Databricks-imposed 0-rate-limit restriction.
Let me know if this helps clear things up or if you need help with the next steps. Good luck!