Hi 浩司 渡部
You're encountering deployment issues and seeing a negative quota value after the release of the o3-pro model in Azure OpenAI. This is a known issue that has affected several users due to recent changes in how Azure manages quotas for high-compute models like o3-pro.
As of August 2024, Azure transitioned from model-specific quotas to a model-independent quota system. This means that instead of having separate quotas for each model (e.g., o3-pro, gpt-4), all deployments now draw from a shared pool of Provisioned Throughput Units (PTUs) per subscription and region. If your existing deployments consume more PTUs than your assigned quota, it can result in a negative quota value, which blocks new deployments or edits to existing ones. This change was automatic and applied to all existing customers without increasing their quota limits unless explicitly requested (source).
To resolve this issue, you should first verify your current quota allocation in the Azure AI Foundry portal under the “Quotas” tab. If your quota is insufficient, you can submit a self-service quota increase request directly from the portal. Look for the link next to the quota item labeled “Provisioned Managed Throughput Unit” and specify the PTU-Managed quota type.
Additionally, ensure that any deleted deployments are fully purged, as they may still count against your quota until completely removed. You can do this through the Azure portal or CLI. If you're still seeing issues, double-check that your deployment names and regions are correctly configured, as errors like “DeploymentNotFound” can also occur if the deployment name is incorrect or not yet fully registered
For more detailed guidance on quota management and troubleshooting, refer to the official Azure documentation on Azure OpenAI quotas and limits and the o3-pro model announcement.
Hope it Helps!
Thanks