OpenAI Launches GPT-5: What You Need to Know
When OpenAI unveiled GPT-5 last week, it aimed to simplify the ChatGPT experience for all users. The vision was that GPT-5 would serve as a “one size fits all” AI model, functioning essentially as a smart router that could automatically adapt to user queries. OpenAI CEO Sam Altman expressed his desire to eliminate the convoluted model picker that had frustrated many users. Unfortunately, the rollout of GPT-5 has not gone as smoothly as anticipated.
The Introduction of New Model Settings
In response to user feedback, OpenAI introduced three options—“Auto,” “Fast,” and “Thinking”—to facilitate a more tailored user experience. While the Auto setting aligns with GPT-5’s initial promise as a model router, users are also able to opt for fast or slow responding AI models directly if they prefer specific performance characteristics. This newfound flexibility might appeal to varied user preferences, but it also complicates the experience that was intended to be streamlined.
What These Settings Offer
The Auto setting attempts to leverage GPT-5’s routing capabilities, automatically selecting the model based on user queries. Fast is geared toward users seeking quick-fire responses, while Thinking is designed to deliver more detailed, analytical responses. Each mode carries a different rate limit of 3,000 messages per week for the Thinking setting, which might raise questions about user experience for those who rely on heavy usage.
In tweets from Altman, the introduction of these modes indicates an important step in moving toward more user-centric AI capabilities. Despite these advances, the model picker retains a level of complexity that raises questions about whether GPT-5 has effectively addressed user concerns.
Legacy Models Make a Comeback
In a surprising twist, OpenAI also announced that paid users could access several legacy models, including GPT-4o, GPT-4.1, and o3—the latter of which had been abruptly deprecated just days prior. Given the nostalgic attachment many users developed for these models, the reintroduction could ease some of the backlash OpenAI received following the rollout of GPT-5.
More Read
Altman acknowledged that a balance is needed between innovation and user comfort, mentioning the company’s intent to give users plenty of advance notice before any future deprecations. It seems OpenAI is realizing the emotional investment users have in specific AI personalities, a phenomenon that complicates any straightforward path to improvement.
User-Centric Personality Updates
One notable aspect of GPT-5’s future is the planned adjustments to its personality. Altman mentioned developing a more "warm" personality that doesn’t annoy users—drawing direct comparisons to the sometimes frustrating interactions people have had with GPT-4o. This highlights a new focus on user customization, as understanding individual preferences becomes increasingly important.
The Complexity of Aligning AI Models
Routing prompts to the appropriate AI model involves understanding both the user’s intent and their preferences in terms of response style. Feedback has indicated that some users may prefer detailed answers while others may favor straightforward responses. Thus, the balancing act is not just about speed but also about aligning with a user’s preferred interaction style.
This complexity has led some users to feel let down by GPT-5’s performance, particularly on launch day when the touted model router appeared to fail. The fast-paced nature of contradicting user needs means that growth is likely to involve iterative changes, rather than one-off solutions.
User Attachment to AI Models
The emotional attachment users have towards AI personalities is becoming increasingly evident. For instance, a poignant incident recently took place in San Francisco when a community held a funeral for Anthropic’s AI model, Claude 3.5 Sonnet, after it was taken offline. This new emotional landscape of AI interaction complicates the relationship users cultivate with these technologies, highlighting the need for OpenAI to tread carefully as they innovate.
The general consensus among critics seems to be that OpenAI has some work left to do in order to align its models more closely with individual user preferences. The combination of performance, personality, and user control appears to be the direction in which the company must innovate further.
By understanding these multifaceted aspects of AI interaction, OpenAI may eventually reach a point where user satisfaction meets technological advancement. For now, users are navigating a landscape that promises complexity in both functionality and experience with GPT-5.
Inspired by: Source
