OpenAI is retiring GPT-4o from ChatGPT

Even as frontier AI gets exponentially smarter, letting go of older conversational companions has proven surprisingly hard for many users.

OpenAI has officially announced that on February 13, 2026, it will retire GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini from ChatGPT. These models will depart alongside the previously announced retirement of the early GPT-5 Instant and Thinking versions.

If your daily work runs on OpenAI's API, you can take a breath. OpenAI confirmed that there are no changes to API availability at this time.

The long goodbye to warmth

GPT-4o has had an unusually resilient afterlife. When OpenAI initially deprecated the model during the launch of GPT-5, pushback from Plus and Pro subscribers forced a swift reversal.

Users told the company they needed more time to transition creative workflows, specifically citing a strong preference for GPT-4o's conversational warmth and distinct personality over the clinical precision of newer releases.

This time, the retirement looks permanent. OpenAI stated that the vast majority of ChatGPT traffic has transitioned to GPT-5.2, with only 0.1% of active daily users still manually selecting GPT-4o.

The feedback from that initial outcry helped shape GPT-5.1 and GPT-5.2, which introduced dedicated controls for personality, warmth, and enthusiasm alongside dedicated styles like Friendly.

Less preaching, treating adults like adults

Beyond the model cleanup, OpenAI dropped notable hints about ChatGPT's immediate product direction.

The company acknowledged ongoing user frustrations regarding unnecessary refusals, overly cautious filters, and preachy responses, promising targeted updates in the near future.

OpenAI also confirmed ongoing development of a ChatGPT experience tailored specifically for adults over 18. Built around the concept of treating adults like adults, the experience aims to grant wider creative and conversational latitude within sensible guardrails.

To lay the groundwork for that adult tier, OpenAI has already deployed automated age-prediction systems for users under 18 across most major global markets.

Why it matters

Retiring GPT-4o closes the book on an era defined by accidental emotional attachment to specific model checkpoints.

For everyday users, clinging to legacy models for creative ideation is no longer an option. If you relied on GPT-4o's specific demeanor, the move now is to actively dial in the personality and warmth sliders in GPT-5.2. It takes extra setup, but it prevents sudden workflow disruption when older weights are phased out.

For the broader AI ecosystem, OpenAI's open acknowledgment of preachy behavior and unnecessary refusals is a welcome course correction. Over the past year, safety alignment often drifted into patronizing lectures, alienating power users and creative writers. Moving toward an adult mode with real customization will likely force competitors to rethink their own heavy-handed moderation layers.

We think OpenAI handled this transition far better than its first abrupt attempt, letting daily adoption drop to a fraction of a percent before pulling the plug. The lingering risk, however, sits squarely in the new age-prediction layer. Relying on behavioral heuristics to determine who gets adult creative freedom and who gets filtered guardrails could easily trigger false positives and fresh privacy headaches.