OpenAI Phases Out GPT-4o and Other Legacy Models

OpenAI announced official discontinuation of GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT — following previously announced GPT-5 (Instant/Thinking) discontinuation, representing a major cleanup of the 4th generation model line. APIs see no immediate changes, but consumer-facing consolidation and flagship concentration are accelerating. This is not simple legacy cleanup but a portfolio strategy transformation redesigning usage patterns, brand tone, operational costs, and safety policies together. Usage transition context: OpenAI disclosed GPT-4o selection rate is 0.1% on daily basis — core usage has already migrated to GPT-5.2. Three strategic meanings of model reduction: (1) Operational efficiency — more models mean distributed infrastructure, evaluation, safety policy, tuning, documentation costs; reducing consumer models simplifies inference cost management and quality control; maintaining API while cleaning up ChatGPT minimizes enterprise transition costs; (2) Brand consistency — different models with different tones and personalities weaken product identity; concentrating on GPT-5.2 enables designing a consistent "what kind of ChatGPT?" experience, affecting long-term NPS and resubscription rates; (3) Safety/policy update acceleration — OpenAI explicitly stated improving unnecessary refusals and excessive moralizing; policy and guardrails become easier to uniformly reflect with fewer models; adult (18+) design version pursuit and age prediction introduction fit the same context. The GPT-4o precedent: GPT-4o was revived once after discontinuation because some users preferred it for creative ideation and "conversational warmth" — this feedback was incorporated into GPT-5.1/5.2 with style/tone selection, warmth/enthusiasm adjustment, and creative idea support enhancement. The lesson: model retirement can only succeed after successfully migrating the capabilities users valued, not just matching benchmark performance.