OpenAI plans to retire GPT-4, launched in March 2023, from ChatGPT on April 30, to be “fully replaced” by GPT-4o; GPT-4 will remain available via OpenAI's API
OpenAI will soon retire GPT-4, an AI model it launched over two years ago, from ChatGPT, according to a changelog posted on Thursday.
Context & Ripple Effects
This is an early step in OpenAI separating the consumer ChatGPT model lineup from API availability: GPT-4 leaves the chat product while remaining accessible to developers. Days later, OpenAI also moved to wind down GPT-4.5 API access, showing that model retirement decisions can differ by distribution channel.
Later coverage shows that replacing familiar models became a recurring product-management challenge: OpenAI signaled further ChatGPT model deprecations, then restored GPT-4o access for paid users after user pushback.
First-order effects
- ChatGPT users are moved from GPT-4 to GPT-4o on April 30, ending direct selection of the older model in the consumer product.
- Developers using GPT-4 through the API are not forced to migrate by this change, preserving an existing production option while ChatGPT standardizes on GPT-4o.
Second-order effects
- The split availability creates different migration timelines for consumer users and API customers, making OpenAI's product-change communication more important for teams that rely on model-specific behavior.
- Keeping GPT-4 in the API while removing it from ChatGPT lets OpenAI simplify the consumer interface without immediately disrupting developer integrations; the later GPT-4.5 API wind-down indicates that this reprieve may be model-specific rather than permanent.
Third-order effects
- If this pattern continues, frontier-model providers will increasingly treat consumer model choice as a managed, rotating service rather than a permanent feature, while APIs become the more explicit compatibility surface.
- Repeated retirements and restorations suggest model lifecycle governance—notice periods, fallback access, and migration paths—will become a competitive differentiator alongside raw model capability.
The trend: This is one data point in the industrialization of AI model portfolios, where providers continuously consolidate consumer offerings while managing developer compatibility separately.