When will the next version for the TimeGEN-1 forecasting model from Nixtla get published to the Foundry?

Question

When will the next version for the TimeGEN-1 forecasting model from Nixtla get published to the Foundry?

Brandon Tempel 20

TimeGEN-1 forecasting AI Model from Nixtla in the Foundry is not the latest version of the model. It sounds like they've been working on it for a while, so do we know when Microsoft will have the latest version available?

0 comments

Answer accepted by question author

Pavankumar Purilla 11,655 Moderator

Hi Brandon Tempel,
Currently, the TimeGEN-1 forecasting model from Nixtla available in Azure AI Foundry is the latest version published in the platform, and Microsoft has not announced any specific timeline for when an updated version will be released. While Nixtla continues to work on advancements to their models, there is no confirmed date for when a newer version will be made available in Foundry. We recommend monitoring the Azure AI Model Catalog, the official Microsoft announcements, and the What's new in Azure OpenAI in Azure AI Foundry Models for the most up-to-date information on new model releases and feature updates.

Pavankumar Purilla 11,655 Reputation points Moderator

2025-08-13T03:30:09.4066667+00:00

Hi Brandon Tempel,
Did you get any chance to check the response. Thank you!
Brandon Tempel 20 Reputation points

2025-08-13T16:22:05.6366667+00:00

What I was trying to understand is why isn't the latest model published to the foundry, or how Nixtla can work with someone to get it updated? I don't fully understand how these partner relationships work, but we reached out to Nixtla and our understanding is they've upgraded the compression algorithm starting in 0.6.5 that caused backward incompatibility with older client versions. If we could use 0.6.6 which includes finetune endpoint capabilities that we are trying to leverage, we could continue our research. The context of the response from Nixtla was they're looking for who/how to get their latest model published so I'm trying this community as well as our EA rep to attempt getting things connected.

1 additional answer

Your answer

Pavankumar Purilla 11,655 Reputation points Moderator

2025-08-13T03:30:09.4066667+00:00

Hi Brandon Tempel,
Did you get any chance to check the response. Thank you!
Brandon Tempel 20 Reputation points

2025-08-13T16:22:05.6366667+00:00

What I was trying to understand is why isn't the latest model published to the foundry, or how Nixtla can work with someone to get it updated? I don't fully understand how these partner relationships work, but we reached out to Nixtla and our understanding is they've upgraded the compression algorithm starting in 0.6.5 that caused backward incompatibility with older client versions. If we could use 0.6.6 which includes finetune endpoint capabilities that we are trying to leverage, we could continue our research. The context of the response from Nixtla was they're looking for who/how to get their latest model published so I'm trying this community as well as our EA rep to attempt getting things connected.

Answer 1

Brandon Tempel 20

What I was trying to understand is why isn't the latest model published to the foundry, or how Nixtla can work with someone to get it updated? I don't fully understand how these partner relationships work, but we reached out to Nixtla and our understanding is they've upgraded the compression algorithm starting in 0.6.5 that caused backward incompatibility with older client versions. If we could use 0.6.6 which includes finetune endpoint capabilities that we are trying to leverage, we could continue our research. The context of the response from Nixtla was they're looking for who/how to get their latest model published so I'm trying this community as well as our EA rep to attempt getting things connected.

Pavankumar Purilla 11,655 Reputation points Moderator

2025-08-14T05:25:24.7866667+00:00

Hi Brandon Tempel,
The version of the Nixtla TimeGEN-1 model currently available in Azure AI Foundry reflects the latest release that has completed Microsoft’s integration, validation, and compliance processes in collaboration with Nixtla. When Nixtla releases a new version, such as 0.6.5 or 0.6.6 with updated capabilities like the finetune endpoint, it cannot be published directly to Foundry until the model provider works with Microsoft’s engineering team to submit the updated version and ensure it meets compatibility, security, and performance requirements for the platform. Any backward-incompatible changes, such as the compression algorithm update in 0.6.5, require additional testing and coordination to maintain reliability for all Foundry users. We recommend that Nixtla continue engaging with their Microsoft partner account manager or engineering contact to progress the publishing of the latest version. Customers can track updates by monitoring the What's new in Azure OpenAI in Azure AI Foundry Models and the model catalog for announcements.