In a move that signals the end of an era for one of its most popular AI personalities, OpenAI officially announced on January 29, 2026, that it will retire GPT-4o and several other legacy models from ChatGPT.
Effective February 13, 2026, the “Omni” model—famed for its human-like warmth and conversational “EQ”—will be removed from the web and mobile interfaces. This decision comes as OpenAI consolidates its user base around the newer, more powerful GPT-5.2 series.
1. The Models Facing Retirement
The “Valentine’s Eve” cleanup isn’t limited to just GPT-4o. OpenAI is clearing out its entire “GPT-4” and early “GPT-5” experimental lineup:
- GPT-4o: The flagship multimodal model launched in May 2024.
- GPT-4.1 & 4.1 mini: The specialized coding-focused variants.
- OpenAI o4-mini: The early “small reasoning” model.
- GPT-5 (Instant & Thinking): The initial “preview” versions of the fifth-generation series.
2. Why Retire a “Fan Favorite”?
OpenAI’s decision is based on a stark reality of user data. Despite the vocal “Save GPT-4o” campaigns on social media, the actual usage has withered.
- The 0.1% Threshold: OpenAI revealed that only 0.1% of daily users still manually select GPT-4o. The remaining 99.9% have shifted to GPT-5.2, which is now the default “router” model for all tasks.
- Maintenance Overhead: Supporting multiple model architectures “distracts from efforts to improve the models most people use,” according to OpenAI’s official blog post.
- The “Warmth” Problem Solved: The main reason users stayed with GPT-4o was its “warmth” compared to the early, “colder” GPT-5 models. With the launch of GPT-5.2’s “Personality Controls,” users can now dial in “Friendly” or “Warm” tones, making GPT-4o’s unique selling point redundant.
3. What Happens on February 13?
If you are one of the few still relying on these legacy models, here is how the transition will work:
- Automatic Migration: All existing conversations that were set to GPT-4o will automatically default to GPT-5.2 for new messages.
- Historical Access: You can still read and review your past chats with GPT-4o, but you won’t be able to “reply” using the retired model.
- API Exception: Developers using GPT-4o through the OpenAI API are safe for now. OpenAI stated there are no changes to API access at this time, providing businesses with a longer runway to migrate their apps.
- Voice Mode: Interestingly, the “ChatGPT Voice” model (which shares a base with 4o) is not retiring. It will continue to operate as a separate, unified experience.
4. The Rise of “Prism” and GPT-5.2
The retirement of GPT-4o clears the stage for OpenAI’s new “Pro” ecosystem. Just days before this announcement, OpenAI launched Prism (prism.openai.com), a dedicated scientific writing and research environment powered exclusively by GPT-5.2.
By retiring the older models, OpenAI is freeing up the compute power needed to support Prism’s complex “deep reasoning” tasks and its real-time whiteboard-to-LaTeX conversion features.
Conclusion: A Necessary Evolution
While many users will miss the specific “vibe” of GPT-4o, its retirement marks the official start of the GPT-5 Era. For OpenAI, 2026 is the year of consolidation—moving away from a fragmented list of “Legacy” and “Beta” models toward a single, customizable intelligence that can be “Friendly” for a casual chat or “Strict” for a scientific paper.
