OpenAI just released its latest language model, GPT-5, looking to make user interaction with the model even simpler with the help of exciting new features. The new model introduces three distinct modes: “Auto,” “Fast,” and “Thinking.” Most users will almost certainly use the “Auto” mode, which lets you automatically select the most appropriate response strategy. When launch day rolled around, problems with the model router’s efficiency had already brought it into question.
OpenAI seems to have engineered GPT-5 to be just that kind of “one size fits all” AI. The model router is meant to help in quick decision-making on AI model to use in answering user queries. On launch day, users immediately felt the pain of a largely useless model router. For many of them, this led to deep frustration.
Sam Altman, the CEO of OpenAI, has said he’s excited for these new capabilities with GPT-5. He envisions it creating a significantly smoother ChatGPT experience for users. He underscored the markedly increased flexibility to be found within the model picker. Although the majority of users will probably use the default setting of “Auto,” others will appreciate some of the unique replies produced by the “Fast” and “Thinking” modes.
“Updates to ChatGPT: You can now choose between ‘Auto’, ‘Fast’, and ‘Thinking’ for GPT-5. Most users will want Auto, but the additional control will be useful for some people.” – Sam Altman
Their addition is a strong signal that OpenAI is serious about continuing to allow for customized, personalized, creative, and novel interactions with AI. The “Fast” mode gives you the quickest possible responses. At the same time, the new “Thinking” mode offers access to more complex analysis, which limits usage to 3,000 messages per week. Users searching for additional capacity can choose to unlock advanced tools during the “Thinking” mode.
Even with those improvements, OpenAI was met with swift criticism for the issues still present in the model router. Users were unimpressed by what they hoped would be a seamless experience, but instead met complications introduced that lowered their overall satisfaction. Nick Turley, a senior associate on the OpenAI team, admitted as much in a public statement.
“We’re not always going to get everything on try #1 but I am very proud of how quickly the team can iterate.” – Nick Turley
Beyond what’s new in GPT-5, paid subscribers still have access to legacy models such as GPT-4o, GPT-4.1 and o3. This translates to buyers having the ability to select from the familiar features of previous generation models. While doing so, they can discover the game-changing potential of GPT-5.
And though most users have greeted the improvements with open arms, some aren’t ready to give up their new favorite AI systems just yet. A similar attachment was on display at a recent EV event. Hundreds of people took symbolic TSAF. When Anthropic announced Claude 3.5 Sonnet’s retirement, a digital burial drew over 500 participants. These kinds of displays reveal the emotional attachment users have come to develop with certain AI technologies.
OpenAI is continuing to hone GPT-5 and address early worries. TMR already has users ecstatic and clamoring to see how well the new model router will perform. The company is clearly devoted to improving UI/UX and feature-set but recognizes that refinement is an iterative process.