Error in AI Foundry playground using Llama-3.3

Hello @Mark Pearson ,

Please let us know if the following response helps-

Thank for the additional validation details. Based on the observed behavior, the current error, and the published model availability guidance, the issue appears to be related to a mismatch between deployment visibility in the catalog and the published serving availability for this model.

The deployment is created successfully and remains selectable in the portal, which shows that the deployment request is accepted at the configuration level. However, the runtime request fails with No callers satisfying location tag 'eu' found for model: llama-3.3-70b-instruct

This error indicates that the request is failing during EU routing resolution, before the model begins processing. The published deployment-type guidance explains that Data Zone requests are processed only within the selected EU or US data zone, and also notes that not all models support all deployment types. In parallel, the published model availability page for Llama‑3.3‑70B‑Instruct currently shows Global standard and Global provisioned deployment paths for this model.

Taken together, this explains the observed behavior:

• The model can appear deployable in the catalog
• The deployment can complete successfully
• Runtime inference can still fail if no eligible EU Data Zone backend is available for that model path

This also explains why the issue is seen in both Playground and REST API. Both paths use the same backend routing layer, so the failure is not limited to the portal experience. The comparison with other working Data Zone Standard models in the same region is also important, because it shows that the broader region, project, authentication, and general Data Zone setup are functioning as expected. The issue therefore appears to be specific to Llama‑3.3‑70B‑Instruct on the EU Data Zone path rather than a general configuration problem.

Please check if the following steps help-

1. Please avoid repeated redeployment of the same configuration Recreating the same deployment is unlikely to change the outcome if the backend route is not currently available
2. Using a temporary workaround if needed
3. Continue with another model already validated under Data Zone Standard
   1. If requirements later allow it, consider Global Standard for this model, since that is the currently published deployment path

Thank you

Hello @Mark Pearson ,

Welcome to Microsoft Q&A .Thank you for reaching out to us.

The deployment of Llama-3.3-70B-Instruct appears to have completed successfully and is selectable in Azure AI Foundry. However, the request is failing in Playground with the following message -No callers satisfying location tag 'eu' found for model: llama-3.3-70b-instruct

Based on the error text and the currently published deployment guidance, the most likely explanation is a routing compatibility issue related to the current deployment type.

he model documentation currently lists Llama-3.3-70B-Instruct under Global Standard and Global Provisioned deployment paths, while the deployment-type guidance also notes that not all models support all deployment types. Because the error specifically references the EU location tag (eu), the failure appears to be happening during backend routing, before the request reaches the model endpoint.

Please check if the following validation steps help-

1. Redeploying the model using Global Standard
   1. Create a new deployment of Llama-3.3-70B-Instruct
   2. Select Global Standard as the deployment type
   3. Retest the deployment in Playground
2. Confirming whether the issue is limited to Playground
   1. Test the same deployment through REST API or SDK
   2. If the same error appears outside Playground, that would confirm the issue is broader than the Playground experience and is likely tied to deployment routing itself
3. Validating whether the issue is model-specific If another deployed model in the same Foundry project works successfully, that would help confirm that the behavior is specific to this deployment path for Llama-3.3-70B-Instruct, rather than a broader project or region issue. This is also a useful comparison point if escalation becomes necessary

At this time, the issue does not look like a prompt, permission, approval, or authentication problem. The error instead points to a failure in locating an eligible serving backend for the request under the current EU routing constraint.If business requirements require EU-constrained processing, the next step would be to validate a model that is explicitly supported for that deployment path. If business requirements allow it, Global Standard is the recommended immediate workaround.

The following references might be helpful , please check them out

• Foundry Models sold by Azure - Microsoft Foundry | Microsoft Learn
• Understanding deployment types in Microsoft Foundry Models - Microsoft Foundry | Microsoft Learn
• Deploy Microsoft Foundry Models in the Foundry portal - Microsoft Foundry | Microsoft Learn
• Microsoft Foundry Playgrounds - Microsoft Foundry | Microsoft Learn

Please let us know if the response was helpful

Thank you