GPT-4o and older models were removed from ChatGPT
OpenAI removed GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT's model picker on February 13, 2026. The models live on in the API — but for the users who built daily habits around 4o's personality, the app they paid for changed underneath them.
- Announced
- February 6, 2026
- Effective
- February 13, 2026
OpenAI announced the retirement on February 6, 2026 and completed it a week later. By OpenAI's own figure, about 0.1% of users still selected GPT-4o daily — but that minority was intensely attached, and the backlash (documented by TechCrunch; see r/4oforever) included thousands of protest messages during a live podcast with OpenAI's CEO.
OpenAI says GPT-5.1/5.2's personality work was shaped by 4o feedback; the displaced cohort largely disagrees, experiencing stronger guardrails and a changed voice as a forced downgrade.
- The specific model personality and response style years of prompts and habits were tuned against.
- Workflows that relied on 4o-era behaviors now run on a model that responds differently to the same prompts.
- The only official escape hatch (API access) is impractical for non-technical users.
The vendor’s path: Use GPT-5.2 in ChatGPT, or the retired models via the API.
The exit lanes, mapped — including what doesn’t carry over.
Each lane below is covered by the Feature Parity Map: every workflow mapped to its replacement, a per-person playbook, adoption verified. Free during beta.
The most common destination for users who valued conversational warmth — and your custom instructions, projects, and tone setup all have mappable equivalents.
For Google Workspace households — your setup maps to Gems and Workspace connectors.
Not sure which way to go — or moving more than one tool? Run the free audit and see your whole stack’s overlap first.