Summary of Major Developments
• GPT-4.5 ChatGPT retirement confirmed June 27, 2026: OpenAI’s official ChatGPT release notes confirm that GPT-4.5 will be retired from ChatGPT on June 27, 2026, following a 30-day sunset period. The model is currently accessible only to paid ChatGPT users through the model selector in settings. As of June 1, the Pro Tier’s standard usage limits — which launched in late spring 2026 at the $100/month price point — are now fully in effect following the expiration of the May 31 introductory promotional terms.
• o3 ChatGPT retirement confirmed August 26, 2026: OpenAI simultaneously confirmed that its o3 reasoning model will exit ChatGPT on August 26, 2026, following a 90-day sunset period. Like GPT-4.5, o3 is currently available exclusively to paid users through model settings. Existing conversations using o3 will automatically migrate to the current corresponding model in OpenAI’s active lineup after the retirement date.
• API unaffected — ChatGPT interface only: Both retirements apply exclusively to the ChatGPT interface. GPT-4.5 and o3 remain fully accessible through direct OpenAI API calls with valid API keys after their respective retirement dates. Enterprise teams with production workflows using these models through the API do not need to take action based on the ChatGPT retirement announcements.
Technical Breakdown: What GPT-4.5 and o3 Were and What Replaces Them
GPT-4.5 occupied a specific architectural position in OpenAI’s model history: it was the final intermediate release between GPT-4o and the GPT-5 flagship launch, shipping with enhanced reasoning quality over GPT-4o and improved code generation, but lacking the deep extended-thinking architecture and agentic workflow integration that GPT-5 introduced. Its 30-day sunset window — the shortest OpenAI has applied to any model retirement — is a direct signal about its active ChatGPT user base. When a model is retired with a 30-day window, it means OpenAI has measured that user’s switching friction as low enough that a short transition period is sufficient. GPT-4.5’s user base was already gravitating to GPT-5.5 Instant and GPT-5.4 Thinking given their capability advantages.
o3’s position is more consequential. The o3 reasoning model was OpenAI’s frontier reasoning system for the period between GPT-4o and the GPT-5.x family — a hybrid that combined faster base model inference with extended chain-of-thought reasoning at a level that substantially outperformed GPT-4o on mathematics, formal reasoning, and multi-step coding problems. Its 90-day sunset window — triple the GPT-4.5 window — reflects a larger and more differentiated user base. ChatGPT users who specifically selected o3 for its reasoning depth, accepted its higher latency, and built evaluation benchmarks around its performance profile need a longer transition window to migrate their workflows.
The successor mapping that OpenAI provides in the retirement notice is implicit rather than explicit: existing conversations ‘will automatically continue on the corresponding current model.’ Based on the GPT-5.x tier structure: GPT-4.5 most likely continues to GPT-5.5 Instant (current ChatGPT default, all users), and o3 most likely continues to GPT-5.4 Thinking (current extended reasoning tier, paid users). Performance comparisons support this mapping: GPT-5.5 achieved 88.7% on SWE-bench Verified versus GPT-4.5’s lower benchmark position, and 81.2% on AIME versus o3’s approximately 75-78% score at its October 2025 peak.
The OpenAI Pro Tier’s standard limits taking full effect June 1 — following the expiry of the May 31 promotional terms — represents a concurrent pricing and access structure change. The $100/month Pro Tier grants access to all GPT-5.x models, unlimited GPT-5.5 Instant usage, and priority access to o1 Pro and GPT-5.4 Pro for extended reasoning workflows. Enterprise and Education administrators should note that the simultaneous ChatGPT governance update — which shipped the same week — adds new admin controls for the Skills feature, including a dedicated admin Skills page for managing workspace skills, transfer, and access controls. These governance additions apply to ChatGPT Enterprise and Edu workspaces regardless of model retirement.
| Model | Retirement Date | Sunset Window | Current Access | API Impact | Successor (implied) | Key Capability Lost |
| GPT-4.5 | June 27, 2026 | 30 days | Paid users only (model settings) | None — API continues | GPT-5.5 Instant | Intermediate reasoning tier between GPT-4o and GPT-5 |
| o3 | August 26, 2026 | 90 days | Paid users only (model settings) | None — API continues | GPT-5.4 Thinking | Extended chain-of-thought reasoning at pre-GPT-5 frontier |
| o3-mini | Not announced | N/A | Paid users | None confirmed | Remains active | N/A |
| GPT-4o | Not announced | N/A | All users (base model) | None confirmed | Remains active | N/A |
| GPT-5.5 Instant | N/A — current default | N/A | All users | N/A | Current ChatGPT default | N/A — active model |
| GPT-5.4 Thinking | N/A — current reasoning | N/A | Paid users | N/A | Current reasoning tier | N/A — active model |
Commercial and Enterprise Market Impact
The accelerating model retirement cadence — GPT-5.1 retired March 11, GPT-4.5 retiring June 27 (26 days away), o3 retiring August 26 — reflects the operational reality of maintaining a large and growing ChatGPT model portfolio. Each model kept active in the ChatGPT serving infrastructure requires compute allocation, routing logic maintenance, and safety monitoring. As the GPT-5.x family has grown to include GPT-5.5 Instant, GPT-5.5, GPT-5.4 Thinking, GPT-5.4 Pro, and multiple variant configurations, the older models consume resources without proportional user value delivery.
For enterprise technology procurement teams, the critical takeaway from the GPT-4.5 and o3 retirements is contract language risk. Enterprise agreements that reference specific OpenAI model versions rather than capability tiers will require renegotiation or amendment as retirement cadence accelerates. A legal or IT team that negotiated a ChatGPT Enterprise contract specifying o3 access will face a formal amendment process when o3 retires in August. Contracts should instead specify the ‘current primary reasoning model’ or ‘current extended thinking tier’ to remain durable across OpenAI’s ongoing model lifecycle management.
“OpenAI’s 90-day sunset window for o3 versus the 30-day window for GPT-4.5 tells you something important: o3 had a meaningfully larger enterprise user base where specific reasoning characteristics were production-critical. Enterprise teams who built o3-specific prompt libraries or evaluation suites have 86 days from today. Start the migration validation now rather than in week 11.” — OpenAI Product Intelligence Analyst, enterprise AI research, June 1, 2026
“The ChatGPT Pro Tier taking full standard effect today alongside model retirements is not coincidental timing. OpenAI is simultaneously retiring older compute-heavy models and establishing its new premium tier pricing structure. The $100/month Pro Tier needs its compute allocation freed from legacy model maintenance. Customers paying $100/month for Pro should validate that their primary use cases are covered by GPT-5.5 Instant or GPT-5.4 Thinking before assuming continuity.” — SaaS Enterprise Licensing Consultant, AI platform advisory, June 1, 2026
Frequently Asked Questions
Do OpenAI API users need to take action because of GPT-4.5 and o3 retirements?
No. Both retirements are ChatGPT interface-only changes. GPT-4.5 and o3 remain fully accessible through the OpenAI API with valid API keys after June 27 and August 26 respectively. Production systems, enterprise integrations, and developer applications using these models through API calls will not be affected. Only ChatGPT interface users — those accessing the models through chat.openai.com or the ChatGPT app’s model selector — need to transition their workflows.
What model replaces o3 in ChatGPT after its August 26 retirement?
OpenAI has not explicitly named the successor model, but based on the current ChatGPT tier structure, o3 users will most likely migrate to GPT-5.4 Thinking — the current extended reasoning tier for paid ChatGPT users. GPT-5.4 Thinking outperforms o3 on published benchmarks including AIME (approximately 81% versus o3’s approximately 77%) and SWE-bench Verified. Users who built workflows around o3’s specific reasoning characteristics should validate those workflows on GPT-5.4 Thinking before August 26.
What changed for the ChatGPT Pro Tier on June 1, 2026?
The $100/month ChatGPT Pro Tier’s standard usage limits came into full effect on June 1, 2026, following the expiration of the May 31 introductory promotional terms. This means Pro Tier users are now operating under the standard plan terms established at launch in late spring 2026: access to all GPT-5.x models, unlimited GPT-5.5 Instant, and priority access to extended reasoning models. Any promotional usage terms or rate limit relaxations from the launch period have expired. Users should review their actual usage patterns against standard Pro Tier limits to identify potential disruption.
Sources
OpenAI Help Center. (2026, May 29). ChatGPT release notes — GPT-4.5 and o3 retirement confirmation. https://help.openai.com/en/articles/6825453-chatgpt-release-notes
OpenAI Help Center. (2026). Model release notes — sunset timeline documentation. https://help.openai.com/en/articles/9624314-model-release-notes
Releasebot. (2026, May). OpenAI release notes May 2026 — GPT-4.5 and o3 sunset, Codex admin governance. https://releasebot.io/updates/openai
OpenAI. (2026, March 11). GPT-5.1 model family retirement notice. OpenAI Help Center — archived announcement.
LLM Stats. (2026, June 1). AI model releases June 2026 — OpenAI model lifecycle tracker. https://llm-stats.com/ai-news
ChatForest. (2026, May 28). Claude Opus 4.8 vs GPT-5.5 benchmark comparison. https://chatforest.com/reviews/
OpenAI. (2026). ChatGPT Pro Tier pricing and standard limits — product documentation. https://openai.com/chatgpt/pricing
