Microsoft Foundry retires Azure OpenAI gpt-4o version 2024-05-13

What changed

Microsoft’s Foundry Models retirement schedule lists Azure OpenAI gpt-4o version 2024-05-13 as Deprecated with a retirement date of October 1, 2026. The documented replacement is gpt-5.1.

Other gpt-4o versions remain on later schedules (for example 2024-08-06 and 2024-11-20 show April 14, 2027). This brief covers the October 1, 2026 retirement of the 2024-05-13 version only.

Microsoft’s lifecycle policy states that for Global Standard, Data Zone Standard, and Standard deployment types, Microsoft manages automatic upgrades when a model version is retired. Provisioned deployments are not auto-upgraded and require manual migration to the replacement model.

Who is affected

Azure / Microsoft Foundry deployments and applications pinned to Azure OpenAI gpt-4o version 2024-05-13. Teams on later gpt-4o versions are not covered by this October 1 deadline.

Required action is not identical for every customer:

• Standard-type deployments (Global Standard, Data Zone Standard, Standard): Microsoft may auto-upgrade to the listed replacement on a rolling, region-by-region schedule (subject to deployment versionUpgradeOption settings).
• Provisioned deployments: no automatic upgrade; operators must migrate manually.

What you need to do

Review before October 1, 2026 if you still rely on gpt-4o 2024-05-13:

1. Confirm which model version each Foundry / Azure OpenAI deployment uses.
2. Identify whether each deployment is Standard-type or Provisioned.
3. For Standard-type deployments: review Microsoft’s auto-upgrade schedule and your versionUpgradeOption setting; validate behavior against gpt-5.1 (Microsoft’s listed replacement) even if upgrade is managed by the service.
4. For Provisioned deployments: plan and execute manual migration to gpt-5.1 (or another evaluated replacement) before retirement—auto-upgrade does not apply.
5. Re-run quality, latency, and safety checks against the target model before cutover.

Microsoft documents that retired models stop serving inference (lifecycle policy describes 410 Gone at retirement for Foundry Models). Confirm exact portal/API behavior in your subscription as you plan.

Migration effort

Likely medium overall as an editorial estimate: Standard auto-upgrades still need validation testing; Provisioned paths require explicit migration work. Effort varies by deployment type, region, and how tightly prompts/tools were tuned to gpt-4o 2024-05-13.

Cost impact

Unknown. Microsoft lists a replacement model but the schedule page does not quantify price differences for equivalent workloads. Compare current Foundry pricing for gpt-4o and gpt-5.1 against your token mix before assuming spend direction.