Key Points
- OpenAI notified developers that five GPT-4 series API models shut down on January 14, 2027.
- The list covers GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini release versions.
- Teams running production code on those models have roughly three months to migrate and test.
The latest:
Five GPT-4 series models lose API access on January 14, 2027, according to OpenAI’s deprecations page, which logged the developer notification on October 14, 2026. The company recommends a single replacement for all of them, gpt-6-luna. Fine-tuned versions built on those models follow the same shutdown date as their base models.
Details:
- The models: According to OpenAI’s deprecations page, the affected snapshots are gpt-4o-2024-08-06 and gpt-4o-2024-11-20, gpt-4o-mini-2024-07-18, gpt-4.1-2025-04-14 and gpt-4.1-mini-2025-04-14. All five carry the same shutdown date, and all five point to gpt-6-luna as the recommended successor.
- Fine-tunes included: The page states that “fine-tuned versions of these models follow the same deprecation schedule as their base models.” That removes any grace period for teams that invested in custom training runs on top of GPT-4o or GPT-4.1, and folds them into the January 14, 2027 cutoff.
- Next wave: A separate October 1, 2026 entry sets April 1, 2027 as the removal date for GPT-5.3-Codex, GPT-5.1 and GPT-5.4-Nano. OpenAI recommends gpt-6-sol as the replacement for GPT-5.3-Codex and GPT-5.1, and gpt-6-luna for GPT-5.4-Nano.
- Speech models: Also dated October 1, 2026, OpenAI said its text-to-speech models leave the API on January 6, 2027, with gpt-realtime-2.1-mini listed as the single replacement for all of them — an earlier deadline than the GPT-4 series cutoff.
- Transcription endpoint: An August 26, 2026 entry gives transcription models a February 26, 2027 removal date, with gpt-live-transcribe or gpt-transcribe as alternatives. The /v1/audio/translations endpoint is being retired in favor of v1/realtime/translations, a routing change rather than a straight model swap.
- Already gone in ChatGPT: OpenAI’s help page on retiring GPT-4o says the consumer product dropped GPT-4o, GPT-4.1, GPT-4.1 mini, o4-mini and GPT-5 Instant and Thinking on February 13, 2026. Enterprise workspaces kept GPT-5 until February 19, 2026.
- Business holdout: Business, Enterprise and Edu customers retained GPT-4o inside custom GPTs until April 3, 2026, after which it was discontinued across all plans. Existing chats and projects defaulted to GPT-5.3 Instant and GPT-5.4 Thinking and Pro equivalents, with custom GPTs moved automatically to the nearest match.
- Unchanged surfaces: The help page says ChatGPT Voice and ChatGPT Images were not affected, despite running on an underlying model similar to GPT-4o. It also said retired ChatGPT models would remain available through the API, with advance notice before any future API shutdown.
Background:
OpenAI maintains a running deprecations log for developers, listing the notification date, the affected model snapshots, the shutdown date and a recommended replacement for each entry. Consumer retirements in ChatGPT are tracked separately through the company’s help center.
Between the lines:
The ChatGPT help page’s promise that retired models stay reachable through the API is now being spent down: GPT-4o left the consumer product in February 2026 and leaves the API in January 2027, roughly eleven months later. With text-to-speech going on January 6, GPT-4 series on January 14 and transcription on February 26, three separate migrations land inside eight weeks.
What’s next
January 6, 2027 for text-to-speech removal, January 14 for the GPT-4 series shutdown, February 26 for transcription models, and April 1 for GPT-5.1, GPT-5.3-Codex and GPT-5.4-Nano.
