3 days ago
7
ZDNET's key takeaways
- GPT-5 now offers four different modes.
- You can still choose Auto to let ChatGPT determine how to respond.
- New choices add complexity to a model that was supposed to be simple.
With the recent release of GPT-5, OpenAI touted its new model as a more efficient one that decides on its own how best to respond to your queries. Gone were the array of different modes that forced you to direct the AI based on whether you wanted a quick answer, longer and deeper thinking, or another variation.
Also: A year after Altman said superintelligence was imminent, GPT-5 is all we get?
Well, with all the user complaints about GPT-5, OpenAI has decided that maybe simplicity isn't such a good thing.
GPT-5 gets four new modes
On Tuesday, OpenAI CEO Sam Altman announced via X that GPT-5 has been updated with four different modes.
The Auto mode is still there for people who don't want or need to tell the AI how to respond. But now there's also the Fast mode for quick answers, a Thinking mode for longer thinking, and a Thinking mini mode for quicker thinking. As Altman said, "most users will want Auto, but the additional controls will be useful for some people."
Also: This free GPT-5 feature is flying under the radar - but it's a game changer for me
Debuting almost a week ago, GPT-5 was an attempt by OpenAI to create a model in which you didn't have to choose among a variety of different modes. The company programmed the new model so it would know when to respond quickly and when to think longer. That meant you could just submit your prompt and let ChatGPT handle the rest.
"GPT-5 is a unified system with a smart, efficient model that answers most questions, a deeper reasoning model (GPT-5 thinking) for harder problems, and a real-time router that quickly decides which to use based on conversation type, complexity, tool needs, and your explicit intent," OpenAI said when introducing GPT-5. "The router is continuously trained on real signals, including when users switch models, preference rates for responses, and measured correctness, improving over time."
Stung by severe criticism
But that desire to reduce complexity landed with a thud. Following its launch, GPT-5 has been stung by severe criticism, much of it from people who bemoaned the sudden loss of GPT-4 and found GPT-5 less than capable in comparison. In an interesting twist, some of the complaints seemed emotional, with some mourning the death of GPT-4 as they might mourn a close friend they had grown to trust and rely on.
In response to the criticism, OpenAI reinstated the legacy GPT-4o, GPT-4.1, o4-mini, and o3 models, but only for Plus and Pro subscribers. For free users, GPT-5 is still the only option. And now GPT-5 has the four new modes, but again only for paid users.
Also: GPT-5 bombed my coding tests, but redeemed itself with code analysis
The dilemma now is that ChatGPT has reverted from a simple and unified process to one with multiple choices. Depending on your query, should you use GPT-5, GPT-4o, o4-mini, or o3? And if you choose GPT-5, should you use one of the specific modes or just set it to Auto and hope for the best?
"4o is back in the model picker for all paid users by default," Altman said in his X post. "If we ever do deprecate it, we will give plenty of notice. Paid users also now have a 'Show additional models' toggle in ChatGPT web settings which will add models like o3, 4.1, and GPT-5 Thinking mini. 4.5 is only available to Pro users -- it costs a lot of GPUs."
An update to GPT-5's personality
With all these recent and sudden changes, OpenAI is clearly struggling to figure out what to do with its latest model. Simplicity? Complexity? A balance between the two? That all remains to be determined.
But Altman's last comment in his X post does provide a clue to the future of ChatGPT.
Also: Here are all the GPT-5 updates OpenAI has rolled out since launch
"We are working on an update to GPT-5's personality which should feel warmer than the current personality but not as annoying (to most users) as GPT-4o," Altman said. "However, one learning for us from the past few days is we really just need to get to a world with more per-user customization of model personality."
English (US) ·