OpenAI to retire GPT-4o and older ChatGPT models on 13 February 2026

OpenAI will retire GPT-4o and several older models from ChatGPT on 13 February 2026. The company said the change is aimed at simplifying choices for users and focusing development on newer systems that now account for the vast majority of usage. API availability remains unchanged for now, and OpenAI has committed to giving advance notice before any future API retirements, according to its guidance.

What is changing in ChatGPT

In a post published on 29 January 2026, OpenAI said it will remove GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini from ChatGPT. The update arrives alongside the previously announced retirement of GPT-5 Instant and GPT-5 Thinking in ChatGPT. OpenAI added that it had brought GPT-4o back temporarily during the GPT-5 rollout after feedback from a subset of Plus and Pro users who valued its conversational warmth and creative ideation, and that this learning has informed improvements to newer models.

• Effective date: 13 February 2026.
• Models exiting ChatGPT: GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini, plus the earlier announced GPT-5 Instant and Thinking.
• API status: No change at this time. The company says it will provide notice ahead of any API impact.
• Conversations and GPTs: Existing chats and custom GPTs will default to GPT-5.2 after the change, as per the Help Centre.
• Voice and Images: OpenAI notes that ChatGPT Voice and ChatGPT Images are not changing as part of this update because they use different underlying models.

Why it matters for users and startups

For everyday users, the removal narrows the model picker to newer options that OpenAI positions as stronger at reasoning, with more granular control over tone and personality. According to the company, usage has already shifted heavily to GPT-5.2, and feedback around GPT-4o’s style has been incorporated into GPT-5.1 and GPT-5.2 through presets such as Friendly and controls for warmth and enthusiasm.

For founders and product teams, the immediate impact is operational. Workflows designed around GPT-4o’s voice and ideation style should be validated on GPT-5.2, with prompts tuned to maintain brand tone and user experience. OpenAI said it is also working to reduce unnecessary refusals and overly cautious responses while continuing safety improvements, including protections for users under 18 in most markets.

How will this affect developers and enterprises

Developers using the API are unaffected for now. OpenAI states that the text models being retired from ChatGPT remain accessible via the API and that it will provide advance notice before any API retirement. For enterprises and education customers, the company’s guidance indicates that administrators can enable limited legacy model access during transition periods. Organisations should treat this as temporary, document model versions in use, and plan migrations to recommended successors.

Context and background

OpenAI previously phased out GPT-4 in ChatGPT in favour of GPT-4o, and more recently restored limited access to GPT-4o for paid users following user feedback. The latest retirement formalises the shift to newer models that aim to combine performance and a more customisable conversational experience. OpenAI said that only a small fraction of users continue to select GPT-4o each day, and that consolidating the line-up helps focus engineering resources on features that most people use.

Timeline and next steps

From 13 February 2026, the listed models will no longer be available in ChatGPT. Users and teams who depend on GPT-4o’s specific tone or behaviour should test and update prompts on GPT-5.2, review style controls, and revalidate customer-facing flows such as support scripts or onboarding assistants. For developers, API access is unchanged at present, but it is prudent to monitor deprecation notices, pin model versions where possible, and budget time for regression testing.

While retiring popular models can be disruptive, OpenAI has framed the move as part of a broader effort to improve personality, creativity, and reliability in ChatGPT while strengthening user choice and safeguards. With early planning and prompt calibration, most teams should be able to navigate the change with minimal friction.