Azure Open AI Batch API - Models roadmap

Microsoft has not published a date or commitment for GPT-6 Luna support in the Azure OpenAI Batch API, so production planning should use the models currently listed in the official Batch support matrix rather than assume GPT-6 Batch availability.

Current Azure Batch availability

As of October 30, 2026, the documented Azure OpenAI Batch matrix includes:

• gpt-5.4-mini (2026-03-17)
• gpt-5.4 (2026-03-05)
• gpt-5.1 (2025-11-13)
• gpt-5 (2025-08-07)
• o3, o4-mini, and selected GPT-4.1/GPT-4o models

These are available through Global Batch, while Data Zone Batch availability depends on the model and region. GPT-6 Luna does not currently appear in either Batch table.^1^

GPT-6 Luna being generally available elsewhere in Microsoft Foundry does not establish availability through the Global-Batch or Data Zone Batch deployment types. Batch is a separate deployment capability with its own model-and-region matrix.^2^

There is also no fixed public cadence for adding models to Azure Batch. A Microsoft Q&A response describes Batch support as arriving after validation, scaling, and regional rollout, but gives no commitment for GPT-6.^3^

Recommended continuity plan

1. Select a currently supported successor.
   Evaluate gpt-5.4 or gpt-5.4-mini against your existing Batch workload. Both are currently listed for Global Batch across the documented regions.^1^
2. Use the v1 API for newer Batch models.
   Microsoft notes that some newer models are unavailable through older API versions and recommends the v1 API for access to newer Global Batch models.^1^
3. Create a separate Batch deployment and run representative jobs.
   In Microsoft Foundry, deploy the candidate using Global-Batch or, where supported, Data Zone Batch. Test output quality, token use, JSON/structured-output behavior, latency, and failure handling before switching production traffic.
4. Decouple deployment names from model names.
   Keep a stable application configuration key, such as BATCH_MODEL_DEPLOYMENT, so migration only requires changing the deployment value rather than rewriting each JSONL record:

{"custom_id":"migration-001","method":"POST","url":"/v1/responses","body":{"model":"batch-successor","input":"Representative production request 1"}}
{"custom_id":"migration-002","method":"POST","url":"/v1/responses","body":{"model":"batch-successor","input":"Representative production request 2"}}

5. Track the two authoritative update locations.
   • The Batch support table for deployment-type and regional support.^1^
   • Azure OpenAI “What’s new,” where previous Global Batch additions have been announced.^4^

Retirement considerations

Microsoft recommends testing newer models and updating applications before the existing model’s retirement date.^5^

Do not treat automatic replacement as the primary migration strategy:

• Replacement selection can occur approximately 90–120 days before retirement.
• Availability varies by deployment type and region.
• Provisioned deployments are not automatically upgraded.
• A replacement available for standard inference is not necessarily Batch-enabled.^5^

If GPT-6 Luna becomes Batch-enabled, it should appear in the Batch support matrix with its model version and applicable regions. Until then, there is no supported Azure GPT-6 Luna Batch deployment or published ETA to plan against.

References

1. Getting started with Azure OpenAI batch deployments
2. GPT-6 Astra, Sol, and Luna for production AI agents in Microsoft ...
3. Availability of newer models for batch deployments - Microsoft Q&A
4. What's new in Azure OpenAI in Azure AI Foundry Models
5. Azure OpenAI in Azure AI Foundry Models model deprecations and retirements