OpenAI completed the retirement of its o3 reasoning model from the ChatGPT model picker on August 26, 2026, closing out a 90-day sunset window that began with a deprecation notice on May 28. Existing conversations that were using o3 will automatically continue on a corresponding GPT-5-series model going forward, and OpenAI is steering users toward GPT-5.4 Thinking or o4 for complex reasoning tasks that previously relied on o3. Importantly, this retirement is scoped to the consumer ChatGPT product only: the higher-compute o3-pro variant remains available to ChatGPT Pro, Team, Enterprise, and Edu subscribers, and the o3 family is untouched in the API, where a separate, later removal schedule applies to specific dated snapshots (o3-2025-04-16 and o3-pro-2025-06-10) set for December 11, 2026. For developers and teams that built workflows, prompts, or evaluation harnesses around o3's specific reasoning behavior, this is a reminder that OpenAI's model lifecycle in ChatGPT now moves on a fairly aggressive cadence: a 90-day notice period is generous by industry standards, but automatic routing to a newer model can silently change output style, latency, and cost characteristics for anything built on top of it. Enterprises with compliance or audit requirements around specific model versions should treat these ChatGPT-side deprecations as a trigger to re-validate prompts or fine-tuned system instructions, even though the API remains stable for now. The broader pattern of frontier labs cycling older reasoning models out within months of release also has procurement implications: teams evaluating vendor lock-in risk should weigh how much of their workflow depends on ChatGPT's consumer product surface versus the more stable API contract.