AI & productivity · August 24, 2026 · 3 min read
ChatGPT is retiring the o3 model: GPT-5.6 Sol takes over from August 26
OpenAI is permanently retiring the o3 reasoning model from ChatGPT on August 26, 2026. If you rely on it for harder tasks, you'll be switched to GPT-5.6 Sol automatically – it's worth checking your saved workflows and any conversations still in progress beforehand.
What happened: On August 26, 2026, OpenAI is permanently retiring the o3 reasoning model from ChatGPT – just two days after this article was published. OpenAI flagged this back on May 28, 2026 in its official release notes, and the shutdown follows the 90-day notice period the company typically gives for larger models. The move is part of a broader lineup cleanup: older GPT-4.5 is disappearing from ChatGPT at the same time. According to OpenAI, o3 is being replaced by the newer GPT-5.6 Sol, which has been the default model for everyday questions since August 6, 2026 and also lets you dial the "effort" level up or down on a slider, a feature we covered earlier. The change affects only ChatGPT (web and mobile apps) for paying users who had o3 enabled in their model settings – API access remains in place for now, on its own, later timeline.
What it means for you: If you specifically picked o3 for harder tasks – reviewing a contract, working through a complex formula, multi-step planning – it will disappear from the model list after August 26 and ChatGPT will automatically route you to GPT-5.6 Sol. In practice, that means three things:
- Check your custom GPTs and saved workflows – if you have a company "custom GPT" or a prompt template that names o3 explicitly in its instructions, verify after the switch that the answers still make sense.
- Back up conversations still in progress – in the weeks before the shutdown, users reported o3 answers disappearing after a page reload in long threads, so copy anything important out of ChatGPT beforehand.
- Try the effort slider on GPT-5.6 Sol – most of what you used o3 for because of its deeper reasoning can now be handled by turning up the effort level on Sol instead of switching models.
How to get started:
- Open the model settings in ChatGPT and check whether you're currently using o3 – if so, try the same kind of task directly in GPT-5.6 Sol with a higher effort setting.
- Review any company "custom GPTs" and internal guides that name o3 specifically, and update them to the currently recommended model.
- For any long conversations still in progress with o3, copy the key answers out of ChatGPT before August 26.
Watch out for: This change affects only the ChatGPT app for regular users – if your company or a vendor accesses o3 through the API (for example in a custom in-house tool), the API runs on a different, later schedule and you don't need to change anything right away. If in doubt, check with whoever manages the API integration at your company before touching something that isn't actually ending yet.
Source: OpenAI Help Center
AI news every week
What matters in AI and productivity, in one weekly newsletter.
1 tip a week · no spam · unsubscribe in one click