GPT-4o realtime retirement date: January 20, 2027
gpt-4o-realtime shuts down on January 20, 2027 (announced Jul 20, 2026).
- Provider
- OpenAI
- Model IDs affected
gpt-4o-realtime- Announced
- July 20, 2026
- Shutdown date
- January 20, 2027
- Replacement
- gpt-realtime-2.1
- Platforms
- Dates apply to the OpenAI API. Fine-tuned models follow the same shutdown schedule as their base model. If you use OpenAI models through another platform, check that platform's own retirement schedule.
- Source
- Official OpenAI deprecations page (data last changed 2026-11-02, re-checked daily)
Migrating from GPT-4o realtime
Not yet reviewed. We haven't written model-specific migration notes for GPT-4o realtime yet. Start with the provider's official guidance and the replacement's model page, and use this general checklist:
- Search your code, config and environment variables for
gpt-4o-realtime. - Re-run your evaluation set against gpt-realtime-2.1 before switching traffic.
- Recount tokens with the new model's tokenizer and re-check
max_tokens, context fit and cost. - Check tool calling, structured output and stop sequences on your real prompts.
- Dates apply to the OpenAI API. Fine-tuned models follow the same shutdown schedule as their base model. If you use OpenAI models through another platform, check that platform's own retirement schedule.
Check your code before Jan 20
Also shutting down on January 20, 2027
- GPT-4o audio
gpt-4o-audio - GPT-4o mini audio
gpt-4o-mini-audio - GPT-4o mini realtime
gpt-4o-mini-realtime - GPT-4o mini transcribe
gpt-4o-mini-transcribe-2025-03-20 - gpt-audio
gpt-audio - gpt-audio-mini
gpt-audio-mini - gpt-realtime
gpt-realtime - gpt-realtime-mini
gpt-realtime-mini
All model retirements · All OpenAI deprecations
FAQ
When will GPT-4o realtime be retired?
GPT-4o realtime (gpt-4o-realtime) shuts down on the OpenAI API on January 20, 2027, according to the official OpenAI API deprecations page.
What should I use instead of gpt-4o-realtime?
The recommended replacement is gpt-realtime-2.1.
What happens to gpt-4o-realtime requests after the shutdown date?
After a model is retired, API requests to it return an error. Move traffic to the replacement and test it on your own prompts before the date.
Does this date apply to Amazon Bedrock, Vertex AI or Azure?
Dates apply to the OpenAI API. Fine-tuned models follow the same shutdown schedule as their base model. If you use OpenAI models through another platform, check that platform’s own retirement schedule.