OpenAI deprecation detail

gpt-4o preview audio and realtime models on a May 7, 2026 shutdown path

This page is a decision brief for teams still pinned to preview realtime or audio model names. It keeps the shutdown date, exact replacement mapping, and the cutover checks in one source-linked view.

Current state

This migration now has a fixed date and a fixed destination.
Deprecated
OpenAI schedules shutdown for the current gpt-4o preview realtime and audio model group on May 7, 2026, with replacements pointing to the current realtime and audio families.

Shutdown date

May 7, 2026

Last checked

March 12, 2026

Replacement path

gpt-realtime-1.5, gpt-realtime-mini, gpt-audio-1.5, or gpt-audio-mini

Why this matters now

This is a dated model-name cutoff, not a vague preview warning.

Use these points to decide whether preview model names are still present in production, whether the move is only a model swap, and whether a separate Realtime interface migration is also still pending.

New work should not start on preview audio or realtime names.
OpenAI has already published a fixed shutdown date for this preview model group, so any net-new use of these names adds avoidable migration debt immediately.
Existing preview model users need cutover before May 7, 2026.
The deprecations page lists the current preview realtime and audio models with a shared May 7, 2026 shutdown date and direct replacements in the current model families.
Model replacement and interface migration are related but separate.
This shutdown group is about preview model names. Teams that also still use the Realtime beta interface have a second migration path because OpenAI documents Beta to GA changes separately in the Realtime guide.

Official sources

Check the OpenAI pages behind this migration brief.

This page uses a fixed source set so the migration brief stays auditable instead of drifting into secondhand advice.

Deprecations

OpenAI deprecations

Source of record for the May 7, 2026 shutdown date, the affected preview model names, and the exact replacement models OpenAI recommends.

Open official page
Guide

Realtime guide

Backs the separate Beta to GA interface migration path, which still matters if a team is changing both model names and the Realtime integration surface.

Open official page

Model mapping

Each preview name now has an explicit current-family replacement.

The deprecations page already names the destination model for each active preview audio or realtime row. The migration work is mostly about finding these names everywhere they still exist and replacing them deliberately.

BeforeNowWhat changesSources
gpt-4o-realtime-preview, gpt-4o-realtime-preview-2025-06-03, and gpt-4o-realtime-preview-2024-12-17gpt-realtime-1.5All three preview realtime names are on the same May 7, 2026 shutdown path, and OpenAI points each one to gpt-realtime-1.5 rather than to another preview alias.
gpt-4o-mini-realtime-previewgpt-realtime-miniThe mini realtime preview row has its own direct replacement, so teams should switch to gpt-realtime-mini instead of keeping a preview naming path alive in configuration.
gpt-4o-audio-previewgpt-audio-1.5The primary preview audio row now has a named current-family replacement, which means production cutover should use gpt-audio-1.5 rather than waiting for the preview alias to disappear.
gpt-4o-mini-audio-previewgpt-audio-miniThe mini audio preview row also has a direct destination, so audio workloads on preview names can be migrated without inventing a custom replacement matrix.

Migration checklist

Treat this as a name audit and cutover sequence first.

The order here is meant to find every preview name still in use, move those references to the listed replacements, and separate that work from any remaining beta interface migration.

Step 1

Inventory every preview audio and realtime model name.

Search application code, SDK wrappers, environment settings, prompt config, and admin controls for the exact preview names listed on the deprecations page.

Step 2

Replace each preview row with the listed current-family model.

Use the exact replacement OpenAI publishes for each row rather than letting a wrapper, alias, or stale default guess the destination model at runtime.

Step 3

Split model replacement from Realtime beta migration work.

If the system still uses the beta Realtime interface, schedule that move separately. The Realtime guide treats Beta to GA as its own migration because the interface changes go beyond model naming.

Step 4

Move production traffic before May 7, 2026.

Treat the OpenAI shutdown date as the operational deadline and leave enough runway for staging validation, rollback planning, and controlled rollout before the listed removal date.

Verify before cutover

Do not close this migration until the preview names are gone from production paths.

These checks are the minimum confidence gates before realtime or audio traffic should leave the preview model group behind.

No preview names remain in runtime configuration

Confirm the exact preview model names on the deprecations page no longer appear in environment variables, service defaults, SDK wrappers, or admin controls.

Replacement mapping is explicit for every affected workload

Verify each preview row has been moved to the current model OpenAI recommends, instead of relying on hidden aliasing or one broad replacement assumption for every workload.

Beta interface assumptions are cleared separately if present

If any realtime client still depends on the beta interface, validate that work against the Realtime Beta to GA guidance rather than assuming the model-name swap finishes the migration.

Production audio and realtime flows pass on the new model names

Re-run the production paths that generate speech, handle transcripts, or manage realtime sessions so the preview-name removal is complete before the May 7, 2026 shutdown date arrives.

Continue the site

Keep moving through the decision from here.

Use the groups below to move laterally through the decision, not back out into another doc hunt.

Related pages

Stay in the same decision neighborhood instead of backing out to search.

OpenAI deprecations tracker

Provider-specific hub for OpenAI shutdown dates, migration briefs, and replacement paths.

Open page

Realtime API beta shutdown

Shutdown-date brief for the Realtime beta interface and the current GA migration path.

Open page

Legacy GPT snapshots

Migration brief for dated legacy GPT snapshots and older preview aliases with a fixed cutoff.

Open page

Compare pages

Open the pages that turn this topic into a side-by-side decision.

GPT-5.4 vs GPT-5 mini

Side-by-side comparison of GPT-5.4 and GPT-5 mini across price, fit, and tool pressure.

Open page

Cheapest OpenAI model for extraction

Scenario recommendation page for choosing the cheapest workable OpenAI extraction model.

Open page

Replacement pages

Use the likely substitutes, migration targets, or fallback choices as the next click.

OpenAI web search pricing

Tool-cost brief for web search pricing across standard and preview search paths.

Open page

OpenAI container pricing

Tool-cost brief for code interpreter container runtime and how it stacks with model spend.

Open page

OpenAI API pricing calculator

Interactive calculator for model tokens, hosted tools, and runtime in one estimate.

Open page

Source category pages

Trace the source families behind this page instead of opening random docs in isolation.

Deprecation and migration sources

Official shutdown, migration, and replacement references behind lifecycle pages.

Open page

Guide and API reference sources

Operational guides and API references used by tool-cost, migration, and calculator pages.

Open page

Return

Return to the OpenAI tracker
Go back to the main OpenAI decision surface to compare this preview-model shutdown risk with current cost signals, tool charges, and the separate Realtime beta migration path.