OpenAI Phases Out GPT-4.5 and o3 From ChatGPT
Key insights
- GPT-4.5 exits ChatGPT on June 27 and o3 on August 26, completing OpenAI's consolidation around the GPT-5.x generation.
- Both retiring models remain fully accessible via the API, meaning developers face no disruption to existing integrations.
- OpenAI assigned different sunset windows: 30 days for GPT-4.5 versus 90 days for o3, reflecting o3's deeper entrenchment in user workflows.
Why this matters
OpenAI's decision to retire models only from ChatGPT while preserving API access signals a deliberate split between its consumer and developer strategies, where the consumer product becomes a tightly curated GPT-5.x experience while the API remains a compatibility layer for enterprise and developer workflows. The asymmetric sunset windows (30 days for GPT-4.5, 90 days for o3) reveal OpenAI's internal assessment of which models have deeper user dependency, a data point for anyone building products on top of OpenAI's model stack. For founders and technical leaders, this retirement cadence establishes a pattern: consumer-facing model availability will compress faster than API availability, creating a divergence that affects product planning timelines and the assumption that ChatGPT and the API offer equivalent model access.
Summary
OpenAI is clearing the last pre-GPT-5 models from ChatGPT. GPT-4.5 exits June 27 after a 30-day sunset; o3 follows August 26 after 90 days. Both models stay fully available through the API.
The consumer product consolidates around GPT-5.x, completing a transition that began with GPT-5's launch. ChatGPT subscribers get fewer model-picker options and a default experience built entirely on the newer generation.
Essentially: OpenAI is separating its API business from its consumer product.
- GPT-4.5 retires June 27; o3 retires August 26, both ChatGPT-only
- API access for both models is unaffected by these dates
- These retirements formally end the pre-reasoning-era model line in ChatGPT
OpenAI is betting GPT-5.x handles every use case that previously required model-switching.
Potential risks and opportunities
Risks
- ChatGPT users with workflows built around o3's reasoning capabilities have a hard August 26 deadline with no guaranteed equivalent inside the consumer product and no rollback option after removal
- OpenAI risks losing power users who relied on explicit model-switching if GPT-5.x underperforms on specialized tasks previously handled by GPT-4.5 or o3 in ChatGPT, accelerating migration to competitors
- Enterprise teams running ChatGPT-facing workflows outside the API must complete migration before August 26 under a 90-day window that may be shorter than their internal change management cycles
Opportunities
- Competing AI assistants (Google Gemini, Anthropic Claude.ai) can directly target displaced ChatGPT power users who want access to a wider, selectable model menu that OpenAI's consumer product no longer offers
- OpenAI API partners and resellers gain a concrete selling point: API subscriptions preserve model optionality that the ChatGPT consumer tier has now eliminated
- Model-serving platforms (Together AI, Fireworks AI) can pitch continued GPT-4.5 and o3 API availability as a migration hedge for teams unwilling or unready to move their production workflows to GPT-5.x
What we don't know yet
- Whether ChatGPT Enterprise and Team plan subscribers face the same June 27 and August 26 deadlines or receive extended access windows not mentioned in the announcement
- Which specific GPT-5.x model replaces o3's reasoning-heavy use cases in ChatGPT, given no direct named reasoning-model equivalent has been designated
- No public statement on eventual API retirement dates for GPT-4.5 and o3, leaving long-term API availability undefined beyond the current consumer-only announcement
Originally reported by help.openai.com
Read the original article →Original headline: OpenAI Announces GPT-4.5 Retiring From ChatGPT June 27 and o3 on August 26 — API Unaffected