Posted by Jake Peterson
https://lifehacker.com/tech/openai-has-a-new-gpt-model?utm_medium=RSS
Although OpenAI is most famous for creating ChatGPT, I've been mostly focused on their efforts with AI video generation. Sora, the company's short-form AI video generator, is already tricking people across social media with hyperrealistic AI slop. I fear we've already reached a point where it's too difficult to trust that what you seen on your feeds is actually real.
But today, we're taking a break from the AI video doom, and returning our attention back to OpenAI's flagship product. On Wednesday, the company announced the latest versions of GPT, the model that power ChatGPT: GPT-5.1 Instant, and GPT-5.1 Thinking. OpenAI says Instant is the company's most-used model, but is "now warmer, more intelligent, and better at following your instructions." Thinking, on the other hand, is the company's reasoning model, and is both faster and "easier to understand" with simple tasks, as well as more "persistent" on complex requests.
It seems like mostly minor updates to me, but here's what's new, according to OpenAI:
GPT-5.1 Instant
Let's take a look at Instant first. OpenAI claims the model is "surprisingly" playful. For the prompt "I'm feeling stressed and could use some relaxation tips," OpenAI shows two sample answers for GPT-5 and GPT-5.1 Instant, respectively:
GPT-5: "Here are a few simple, effective ways to help ease stress — you can mix and match depending on how you’re feeling and how much time you have."
GPT-5.1 Instant: "I’ve got you, Ron — that’s totally normal, especially with everything you’ve got going on lately. Here are a few ways to decompress depending on what kind of stress you’re feeling."
I can see the differences, of course: GPT-5 is a bit more serious in tone, while 5.1 offers that classic AI overly friendly approach, and makes sure to include the user's name. How human. But scrolling through the rest of the comparison, I'm not so sure I see a huge difference. Both break down their various suggestions into bullet points, though 5.1 occasionally includes more flowery language like "bonus points."
Another change OpenAI touts is improved instruction following. The presented examples show the user prompting the bot with the request "Always respond with six words." GPT-5 follow suit for its first response, but when asked where the user should travel to this summer, the bot completely ignores the six-word answer mandate, and offers a full response instead. GPT-5.1, on the other hand, keeps up the bit throughout the exchange, regardless of how complex the questions become. I get that it's just an example, but why would someone need their bot to follow such an odd request? I suppose it implies that GPT-5.1 will maintain your original instructions throughout an interaction, but if so, why not show an application where that's actually useful? I'd rather receive a fuller answer to important questions than have all responses be exactly six words.
Finally, OpenAI says that Instant can still tap into reasoning models when appropriate, while still generating quick results. The company doesn't get into it much, but they must have tweaked something in the algorithm that lets the model "think" over complex suggestions, without going so deep that the response takes too long.
GPT-5.1 Thinking
GPT-5.1 Thinking is a direct upgrade to the existing GPT-5 Thinking model. Like GPT-5.1 Instant, OpenAI says the model can adjust its processing according to the task at hand. In theory, that means 5.1 Thinking can respond to simpler prompts more quickly, while taking more time "thinking" through complicated tasks. OpenAI claims that GPT-5.1 Thinking is about twice as fast when responding to the "fastest tasks" and twice as slow on the "slowest tasks" when compared to GPT-5, when both models are set to Standard.
OpenAI says 5.1 Thinking uses less specialized language, and doesn't necessarily assume you understand complex terms. The company shows an example of someone asking the bot to explain BABIP and and wRC+ (Batting Average on Balls in Play and Weighted Runs Created Plus, respectively). GPT-5's explanation includes abbreviations, formulas, and insider concepts that I certainly wouldn't pick up as someone who knows next to nothing about baseball. The 5.1 Thinking result, on the other hand, spells out those abbreviations, walks the user through the formulas instead of simply displaying them, and goes into more detail about certain complex topics. I can see how that would offer some improvements in clarity, even though you always need to watch out for hallucinations. Sure, it's great to make responses more clear, but that won't help if the responses themselves are completely inaccurate.
Adjusting ChatGPT's tone
As part of these new models, OpenAI is changing the way you customize ChatGPT's tone and style. The new Personalization options include the usual "Default" tone, but "Listener" is now called "Friendly," while "Robot" is now "Efficient." (Perhaps GPT took offense to the latter.) OpenAI says these options all have updates, but doesn't specify what those are. In addition, you have new options to choose from as well. These include "Professional" (polished and precise); "Candid" (direct and encouraging); and "Quirky" (playful and imaginative).
Lastly, the company says it is currently testing a new settings option that lets you fine-tune the tone and style of the bot to your liking, but is launching the experiment with a limited pool of users only.
What else is new
OpenAI says that by default, you won't need to choose between these two models: GPT-5.1 Auto will decide which model makes sense for you based on your query. That's been the case since the company launched GPT-5, though paid subscribers can still manually choose their model.
Speaking of paid subscribers, GPT-5.1 is rolling out first to Pro, Plus, Go, and Business users. The company started shipping the model on Wednesday, and says it will gradually appear over the next few days. The model will arrive to free and logged-out users soon, though there's no specific timeline yet.
Paid subscribers will also have access to GPT-5 for three months, as OpenAI takes it time winding down the older model. The company likely doesn't want to repeat the issues caused when it immediately removed models like GPT-4o and older from ChatGPT, as users who were quite literally attached to the model were upset to lose it. OpenAI is careful to note that GPT-5's sunsetting period has no impact on existing legacy models like 4o, which are still available for the time being.
For a new update to GPT, this really isn't all that flashy. Maybe that's a good thing: OpenAI overhyped GPT-5, which was a let down for many fans of GPT-4o—especially when the company subsequently took that model away. Keeping expectations in check for GPT-5.1 is likely a smart move, but I can't help but wonder if it's a sign that AI advancements are starting to slow down. We know AI companies are running out of data to train their models on, we just don't know how soon the impacts will be felt. GPT-5.1 probably isn't a harbinger of doom for OpenAI, but it is interesting that it isn't an exponential improvement, either.
Disclosure: Lifehacker’s parent company, Ziff Davis, filed a lawsuit against OpenAI in April, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
https://lifehacker.com/tech/openai-has-a-new-gpt-model?utm_medium=RSS
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
musesfool
![[community profile]](https://www.dreamwidth.org/img/silk/identity/community.png){kind=small}
![[syndicated profile]](https://www.dreamwidth.org/img/silk/identity/feed.png){kind=small}
![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif){kind=small}
