Source: Build Fast with AI — OpenAI's GPT-4.5 quietly retired from ChatGPT on Friday, June 27, 2026, after a 30-day sunset period. Existing conversations using the model have been automatically migrated to GPT-5.5, while the API continues to support GPT-4.5 separately.
GPT-4.5 launched in February 2026 as OpenAI's final pre-GPT-5 frontier model. It represented the ceiling of what the GPT-4 architecture could achieve before the capability step-change that GPT-5 introduced. For consumers, this marks the end of the GPT-4 era inside ChatGPT's interface — a generation that began with the original GPT-4 release in March 2023 and defined public expectations of what AI language models could do.
The retirement also includes GPT-o3, scheduled for August 26, 2026 after a 90-day sunset. For development teams that built workflows around GPT-4.5's specific reasoning style or voice, the migration means recalibrating against GPT-5.5's different behaviour profile. Citation trails and response patterns built on GPT-4.5 are now frozen in time.
Analysis
GPT-4.5's quiet exit tells you more about OpenAI's product strategy than any launch event could. The company is accelerating model deprecation cycles as aggressively as it ships new ones — the 30-day sunset for GPT-4.5 is half the window it gave GPT-o3, and a fraction of what enterprise customers typically get from infrastructure vendors. This isn't just housekeeping. Forcing consumers onto GPT-5.5 makes the model the training flywheel: every conversation now feeds the reinforcement pipeline for the next tier, and the "corresponding" migration language in the release notes deliberately avoids mentioning whether chat history fidelity survives the transition. The API carve-out is the tell: enterprise customers paying per-token keep their stable endpoints, while the consumer layer is treated as a rolling beta. When GPT-5.6 ships — potentially limited to government-approved partners already — expect this gap between consumer and API access to become formal policy.
