OpenAI has reintroduced control with its latest release, GPT-5. Originally designed as a universal model that automatically picked the best response, it now offers settings like ‘Auto’, ‘Fast’, and ‘Thinking’—letting you choose the balance between speed and thoughtfulness that suits your needs.
CEO Sam Altman, who has long critiqued unwieldy menus, confirmed that subscribers can access previous models such as GPT-4o, GPT-4.1, and o3. For many users, having the option to revert to tried-and-tested choices brings a welcome familiarity.
Altman also shared plans to adjust GPT-5’s personality. The aim is to create a version that’s agreeable without the quirks some found in GPT-4o, reflecting an ongoing effort to personalise AI responses to individual preferences.
Despite these updates, the model picker remains a bit of a maze. Early troubles with the router led to performance hiccups and user complaints, prompting swift responses during a Reddit AMA. OpenAI’s team is clearly working hard to streamline the process.
It’s interesting to note that preferences go beyond mere speed. Some users have grown attached to the distinctive, sometimes contrarian voices of older models—a sentiment recently echoed in San Francisco when hundreds expressed disappointment over the deactivation of Anthropic’s Claude 3.5 Sonnet.
If you’ve ever struggled with finding the right AI tone, you’ll appreciate that future adjustments—and any necessary model retirements—will come with plenty of notice. OpenAI’s commitment to enhancing user choice remains at the forefront of these developments.
