On January 29, 2026, OpenAI used a blog post to announce the end of an era: starting February 13, GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini disappear from ChatGPT, joining the previously announced retirements of GPT-5 Instant and GPT-5 Thinking on the same schedule. Citing OpenAI’s own figures, CNBC reports that only about 0.1% of users still select GPT-4o each day; the vast majority have long since moved to GPT-5.2. The API is not part of this wave.
What makes the story worth a closer look is not that another model is being deprecated, but that ChatGPT’s model menu is collapsing onto a single generation. For product teams, it is a live case study in how suppliers wind down long-tail model preferences: a loyal group of users deeply attached to an old model’s personality, meeting a resource-allocation logic that looks at usage percentages.
What’s Being Retired, and When
The essentials of the announcement:
- Models: GPT-4o (launched May 2024), GPT-4.1, GPT-4.1 mini, and o4-mini, plus the earlier-announced GPT-5 Instant and GPT-5 Thinking
- Effective date: February 13, 2026, when they leave ChatGPT
- Usage: OpenAI says roughly 0.1% of users pick GPT-4o daily
- API: explicitly unaffected — code pinned to these model IDs keeps working
Why OpenAI Says the Time Is Right
OpenAI’s stated reason is that recent improvements to GPT-5’s “personality, customization, and creative ideation” made it comfortable pulling the plug. The official language is unusually candid: “We know that losing access to GPT-4o will feel frustrating for some users, and we didn’t make this decision lightly,” and “Retiring models is never easy, but it allows us to focus on improving the models most people use today.”
The 0.1% figure is the heart of the decision. Against ChatGPT’s user base, that fraction still translates into a substantial number of people, many of them paying subscribers — GPT-4o was beloved for its warm conversational style. But for OpenAI, the cost and complexity of keeping several legacy models alive has clearly outgrown the revenue from that group. SKU consolidation is the shared instinct across frontier labs this quarter; nearly every major supplier is pruning its model lineup in the same direction (see our 2026 opening outlook).
The API Is Not in This Wave
CNBC’s reporting is explicit that the API is untouched, which matters for developers: if your product depends on GPT-4o’s behavior, nothing forces a migration today. The real exposure is experiential. Once the ChatGPT-side option disappears, users who preferred the old personality have nowhere to channel that preference, and their tolerance for the new default will show up directly in feedback and retention. And if your marketing copy still sells “GPT-4o” as a feature, this is the moment to update it.
From the August 2025 Backlash to a “Plenty of Notice” Pledge
The emotional attachment to GPT-4o has a documented history. When GPT-5 launched in August 2025, OpenAI briefly removed GPT-4o access entirely, drew an intense backlash from paying users, restored it within days, and watched CEO Sam Altman pledge on X that any future retirement would come with “plenty of notice.” This time the announcement landed on January 29 with a February 13 effective date — about two weeks of lead time. That is the minimum viable fulfillment of the pledge, and it signals that OpenAI now treats model retirements as product events requiring PR management, not background maintenance.
What It Means for Product and Engineering Teams
Three practical takeaways. First, preferences over model personality are a real long-tail need — at ChatGPT’s scale, 0.1% daily usage is enough to generate a news cycle, so never rely on average usage alone when retiring a beloved default. Second, suppliers consolidating their model menus is now the norm; build products on a stable model family and an abstraction layer, not on one specific legacy checkpoint. Third, the communication rhythm around a retirement is itself a product skill: how long the notice period runs, and whether the replacement’s personality actually matches, will be scrutinized by exactly the users most likely to complain publicly.
Sources
- OpenAI will retire GPT-4o from ChatGPT next month — CNBC
- Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini — OpenAI
- ChatGPT Release Notes — OpenAI Help Center
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.