An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
Hi Rithvik Gabri,
This behavior typically indicates a backend provisioning failure for the Claude model rather than an issue with your code or deployment configuration.
In Azure AI Foundry, Anthropic Claude model deployments are subject to region‑ and subscription‑level capacity and entitlement checks. When provisioning fails, the deployment can enter a terminal Failed state. Once this happens, the deployment can’t be recovered or reused.
A few important points based on current Foundry behavior:
- A
Provisioning state: Failedfor Claude models usually means the service could not allocate capacity for that model in the selected region or subscription. - Deleting and recreating the same deployment or switching between Claude variants (for example Sonnet / Opus / Haiku), typically does not resolve the issue if capacity or entitlement is unavailable.
- This is a platform‑level limitation, not something that can be fixed via SDK, API parameters, or retry logic.
The only supported path today is to:
- Verify that the Claude model is available and enabled for your subscription and region.
- Attempt deployment in another supported region where Claude capacity may be available.
- Monitor Foundry documentation and release notes for updates as Claude capacity and regional support expand.
This limitation has been acknowledged in other Foundry discussions, and any future support for improved provisioning or recovery would be reflected in official Foundry documentation and announcements.
Hope this helps clarify the current behavior. Do let me know if you have any further queries.
If this answers your query, please do click Accept Answer and Yes for was this answer helpful.
Thankyou!