GPT-6.1 Terra available in Microsoft Foundry
- Maxime Hiez
- Foundry
- 02 Nov, 2026
Introduction
GPT-6.1 Terra has been generally available in Microsoft Foundry since October 26, 2026, the same day OpenAI launched it. Two deployment types are offered, Global Standard and Data Zone Standard. The Responses API requires no new SDK.
Check the GPT-6.1 Terra article HERE.
Deployment and regions
In the new Foundry portal, deployment goes through Discover > Models, the model card, then Deploy. The deployment name is the value to pass in the model parameter of the https://resource-name.openai.azure.com/openai/v1/ route, without api-version. The project and the resource must be in a supported region.
gpt-6.1-terra (version 2026-10-26) carries no Preview label : it follows the lifecycle of generally available models. The context window is 1 050 000 tokens, of which 922 000 are input and 128 000 are output.
Global Standard is offered in all supported regions except eastus2 and northeurope; Data Zone Standard covers the United States (without eastus2), Europe (without northeurope) and Asia-Pacific. Canada Central and Canada East are available in Global Standard only.
info
Billing and quotas
Billing is per token, with no Azure Marketplace subscription. Prices from the Microsoft article, in US dollars per 1M tokens :
| Deployment | Context | Input | Output |
|---|---|---|---|
| Global Standard | Short | 1.00$ | 5.00$ |
| Global Standard | Long | 2.00$ | 7.50$ |
| Data Zone Standard (United States) | Short | 1.10$ | 5.50$ |
| Data Zone Standard (United States) | Long | 2.20$ | 8.25$ |
| Data Zone Standard (Europe, Asia-Pacific) | Short | 1.20$ | 6.00$ |
| Data Zone Standard (Europe, Asia-Pacific) | Long | 2.40$ | 9.00$ |
The article also publishes cache prices. The Global rows match the Azure pricing page exactly. Its only Data Zone row repeats the article’s US figures, rounded for the cache, so the European and Asia-Pacific prices rest on the article alone.
Short or long context is measured on the input tokens of a single request. Learn does not state the threshold for Terra.
Default quotas depend on the subscription tier. In Global Standard, they range from 250 requests and 250 000 tokens per minute at tier 1 to 8 000 and 8 000 000 at tier 6 (4 000 and 4 000 000 at tier 4). In Data Zone Standard, they range from 100 and 100 000 to 2 667 and 2 667 000.
From tier 4 onward, they are lower than those of GPT-6.1 Sol (15 000 requests per minute in Global Standard at tier 6) and GPT-5.6 Terra. For Terra, nothing indicates that an access request is required, whereas for Sol the reasoning models page says “No access request needed.” An additional quota request remains possible through a form, and the subscription tier stays the same.
What admins need to know
Azure Policy can block a deployment type, by SKU name GlobalStandard or DataZoneStandard, at the subscription level. It is the lever for a tenant under a residency constraint.
No retirement date for gpt-6.1-terra appears in the deprecation schedule; the lifecycle policy provides for a date set at launch, available through the Models API, and at least 60 days’ notice before retirement. Automatic upgrades are controlled with versionUpgradeOption.
What the model brings
Benchmarks and general capabilities are covered in the launch article. Foundry adds multi-agent orchestration, in preview.
Conclusion
Terra follows the lifecycle of generally available models, which Microsoft defines as production-ready. In Canada, it runs in Global Standard only, with no guarantee that processing stays in the country, and default quotas are lower than Sol’s from tier 4. To check before deploying : the resource region, the tier, and the threshold between short and long context.
Sources
Microsoft Learn - Foundry models sold directly by Azure
Microsoft Learn - Region availability
Microsoft Learn - Quotas and limits
Microsoft Learn - Deployment types
Microsoft Learn - Model lifecycle
Did you enjoy this post ? If you have any questions, comments or suggestions, please feel free to send me a message from the contact form.
Don’t forget to follow us and share this post.