OpenAI retires GPT-4.5 and o3 as IPO talk heats up

OpenAI updated its official ChatGPT release notes this month to confirm retirement timelines for older models and to publish transition guidance for developers and enterprise customers. The company listed specific sunset dates and migration recommendations in a terse product-posting on its help center.

The release notes name GPT‑4.5 as scheduled to be removed from ChatGPT on June 27, 2026, following a 30‑day sunset period, and OpenAI o3 as slated for retirement on August 26, 2026 after a 90‑day sunset window. The company framed the moves as part of ongoing model consolidation.

Those retirements follow earlier waves of model removals announced in February, when OpenAI retired GPT‑4o, GPT‑4.1 and several variants from ChatGPT and adjusted which models are available to paid tiers. The change log highlights that the deprecations specifically affect the ChatGPT product environment and that availability has shifted over recent months.

OpenAI’s platform documentation and deprecation pages lay out technical steps and timelines developers should use to migrate. The API deprecations listing and model release notes explain snapshot timelines, recommended replacement models, and backward‑compatibility caveats for integrations. Teams are urged to check their model snapshots and update defaults in SDKs and frameworks.

For enterprise customers the notices are consequential: some business contracts and internal pipelines still rely on legacy model behavior, and OpenAI’s enterprise pages show paid tiers have had selective access to older models for continuity while transitions proceed. OpenAI told business users how long certain legacy access would persist in custom deployments.

Not every retirement is identical. Coverage of the release notes and migration guides notes that the ChatGPT retirements do not always map one‑for‑one to API removals, and in prior deprecation events OpenAI has staggered API sunset dates separately from ChatGPT product changes. Teams using the API should consult the platform deprecations page for precise removal dates and compatibility notes.

OpenAI framed the clean‑up as a way to focus resources on newer, higher‑capacity models and to improve overall product quality and safety. Outside observers noted the shift reduces fragmentation in the ChatGPT interface while nudging users and partners toward the company’s most capable models. Coverage by industry outlets described the change as quiet but material for heavy users.

The timing of the release‑note update comes amid fresh IPO activity at OpenAI: the company filed a confidential draft S‑1 in early June 2026, a move widely reported by major news organizations and financial outlets. That filing has prompted analysts and reporters to track product and revenue changes more closely.

Market and product watchers say product housekeeping like model retirements can be part of preparing for a public listing, since streamlining offerings and clarifying enterprise support helps to present cleaner revenue and risk profiles to investors. Reporting on OpenAI’s confidential filing links the timing of product and commercial moves to broader capitalization plans.

Practically, developers should inventory usage of GPT‑4.5 and o3 in both ChatGPT‑based workflows and API calls, run regression tests against recommended replacements, and update any orchestration or prompting layers that rely on legacy model outputs. Third‑party tooling and SDKs — from LangChain to LlamaIndex and others — are already shifting defaults as upstream models are deprecated.

The community response has been mixed. Some teams report friction when a familiar model is removed, especially where subtle response‑style differences affect product behavior; other developers welcome a simpler set of supported models that reduce maintenance burden. Industry coverage and user forums captured both annoyance and acceptance.

OpenAI’s product pages and newsroom updates indicate the company will continue to publish release notes and deprecation schedules as changes roll out. Enterprises and integrators should treat those pages as primary sources and set internal timelines that match OpenAI’s sunset windows to avoid outages.

As OpenAI heads toward a potential public listing, this round of retirements is the type of operational tightening investors will watch: it speaks to product focus, cost allocation and how the company manages enterprise continuity. For now, the concrete steps for teams are straightforward — audit, test, migrate — even if the corporate backdrop is evolving quickly.