Non OpenAI model (DeepSeek-V3-0324, etc) with Azure AI Agents gives "invalid_engine_error" error in AI Foundry and Python SDK

Question

Non OpenAI model (DeepSeek-V3-0324, etc) with Azure AI Agents gives "invalid_engine_error" error in AI Foundry and Python SDK

Bhumika Korat 20

I have created a project in AI Foundry for agent usage. I have an Azure AI service connection where I have deployed DeepSeek-V3-0324 model, which is in eligibility list for Agents usage. I have the AI service resource connected with the hub and the project as well.

When I try this model in "Chat Playground" It works. But when I try the "Agent Playground", I get invalid_engine_error: No connection matching model: saeideepseekv3 . From Python SDK, I get RunStatus.FAILED for create_stream as well as create_and_process_run

I have the service connections checked against the steps in the documentations. What am I missing here?

Bhumika Korat 20 Reputation points

2025-04-24T14:41:05.9766667+00:00

FYI: I am facing the same issue with mistral-small-2503. But any gpt model in the same Azure AI Service (Not open ai service), is working with agents.
Bhumika Korat 20 Reputation points

2025-04-25T10:02:35.73+00:00

I also tried o4-mini. It did not work with agents too. Pretty much except gpt, non of the above listed model works and Azure does not even have proper error message to identify issue. llama and cohere is not allowed for me to deploy with weird marketplace subscription error. o3 is not available due to eligibility. Mistral does not work. Coehere command r series is deprecated. While Microsoft is claiming a flexible model switching in ai agents everywhere.
Saideep Anchuri 7,040 Reputation points Microsoft External Staff Moderator

2025-04-28T15:41:04.07+00:00

Hi Bhumika Korat

It seems that you are experiencing issues with deploying non-OpenAI models using Azure AI Agents, specifically receiving an "invalid_engine_error" and encountering various limitations with model availability and marketplace subscriptions.

The Azure AI Agent Service supports a variety of models, but the availability and compatibility may vary by region and specific deployment configurations. For instance, while Azure OpenAI models like GPT-4 are widely supported, non-Microsoft models such as those from the Azure AI Foundry model catalog may have different requirements and limitations.

If you are facing issues with specific models like DeepSeek-V3-0324, o4-mini, or others, it may be due to regional restrictions, eligibility criteria, or deprecated status of certain models. For example, the Cohere command r series has been deprecated, which could lead to deployment issues.

kindly refer below link: model-region-support

legacy-models

Thank You.
Bhumika Korat 20 Reputation points

2025-04-28T15:53:25.1366667+00:00

Thank you for the explanation. I have tested all available "Agent supported" in the model catalogue. I was able to create agent and it only gave error when the request was "run" for the mentioned models. Different error message for different model. I would use "Non-GPT" term for the models I faced issues with.
Additionally, since this morning I started getting "Deployment not found" error specifically in Agent playground for "Non-gpt" models, while the same model deployment works for Chat playground.

Which means Azure has not integrated these models for agents yet, is my assumption correct?
Bhumika Korat 20 Reputation points

2025-04-28T16:05:06.3566667+00:00

Understood. Thank you for the clarification and confirmation.
Saideep Anchuri 7,040 Reputation points Microsoft External Staff Moderator

2025-04-28T16:07:16.6533333+00:00

Hi Bhumika Korat

Thanks for your patience, from the above conversation please tell us if anything was helpful to you, so that we can convert it to answer. Then, if you could Accept Answer and Upvote it for the benefit of community, it will be helpful to others.

Thank you

Accepted answer

0 additional answers

Your answer

Bhumika Korat 20 Reputation points

2025-04-24T14:41:05.9766667+00:00

FYI: I am facing the same issue with mistral-small-2503. But any gpt model in the same Azure AI Service (Not open ai service), is working with agents.
Bhumika Korat 20 Reputation points

2025-04-25T10:02:35.73+00:00

I also tried o4-mini. It did not work with agents too. Pretty much except gpt, non of the above listed model works and Azure does not even have proper error message to identify issue. llama and cohere is not allowed for me to deploy with weird marketplace subscription error. o3 is not available due to eligibility. Mistral does not work. Coehere command r series is deprecated. While Microsoft is claiming a flexible model switching in ai agents everywhere.
Saideep Anchuri 7,040 Reputation points Microsoft External Staff Moderator

2025-04-28T15:41:04.07+00:00

Hi Bhumika Korat

It seems that you are experiencing issues with deploying non-OpenAI models using Azure AI Agents, specifically receiving an "invalid_engine_error" and encountering various limitations with model availability and marketplace subscriptions.

The Azure AI Agent Service supports a variety of models, but the availability and compatibility may vary by region and specific deployment configurations. For instance, while Azure OpenAI models like GPT-4 are widely supported, non-Microsoft models such as those from the Azure AI Foundry model catalog may have different requirements and limitations.

If you are facing issues with specific models like DeepSeek-V3-0324, o4-mini, or others, it may be due to regional restrictions, eligibility criteria, or deprecated status of certain models. For example, the Cohere command r series has been deprecated, which could lead to deployment issues.

kindly refer below link: model-region-support

legacy-models

Thank You.
Bhumika Korat 20 Reputation points

2025-04-28T15:53:25.1366667+00:00

Thank you for the explanation. I have tested all available "Agent supported" in the model catalogue. I was able to create agent and it only gave error when the request was "run" for the mentioned models. Different error message for different model. I would use "Non-GPT" term for the models I faced issues with.
Additionally, since this morning I started getting "Deployment not found" error specifically in Agent playground for "Non-gpt" models, while the same model deployment works for Chat playground.

Which means Azure has not integrated these models for agents yet, is my assumption correct?
Bhumika Korat 20 Reputation points

2025-04-28T16:05:06.3566667+00:00

Understood. Thank you for the clarification and confirmation.
Saideep Anchuri 7,040 Reputation points Microsoft External Staff Moderator

2025-04-28T16:07:16.6533333+00:00

Hi Bhumika Korat

Thanks for your patience, from the above conversation please tell us if anything was helpful to you, so that we can convert it to answer. Then, if you could Accept Answer and Upvote it for the benefit of community, it will be helpful to others.

Thank you

Answer 1

Hi Bhumika Korat

The non-GPT models in Azure AI Agents indicate that these models may not be fully integrated or supported for use with agents at this time. The Azure AI model inference in AI services supports models that have pay-as-you-go billing, but not all models in the Azure AI catalog are supported in Azure AI services. This could explain the discrepancies you're experiencing between the agent playground and the chat playground.

Your assumption that Azure has not integrated these non-GPT models for agents yet aligns with the information regarding model support and deployment options in Azure AI services.

Thank You.

Share via

Non OpenAI model (DeepSeek-V3-0324, etc) with Azure AI Agents gives "invalid_engine_error" error in AI Foundry and Python SDK

0 additional answers

Your answer