Chatgpt gets a great upgrade. Here’s what is new with GPT-5

Expect your Chatgpt Experience to get faster and smarter today.

Openai has updated its main line for Language models Thursday, unveiled the veil GPT-5 Artificial intelligence model yet Months of anticipation. While the developer issued a lot of Form updates In recent months, including the new Open weight models Only this week, more than two years have passed since the appearance of GPT-4. With a new generation that deserves a new number, what is the size of the change you should expect?

“I tried to return to GPT-4 and it was very miserable,” Sam Altman, CEO of Openai, told reporters. “This is much better in clear ways and hidden ways.”

Like its predecessor, GPT-5 operates Chatbots, agents and search tools that you use in ChatgPT or through other applications that use Openai technology. But the company said that this version is more intelligent, more accurate and faster. The demonstrations showed that they were rapidly creating custom applications without coding, and the developers said they worked on ways to ensure that they provide safer answers to questions that could be treacherous. (Disclosure: Zif Davis, the parent company CNET, filed a lawsuit against Openai, claimed that it had violated the copyright of ZifF Davis in training and operating artificial intelligence systems.)

The new model should be available to everyone on Thursday, including those who use the free layer of Chatgpt. This is what to expect.

One model for everyone (somewhat)

Unlike some additional versions of Openai, the GPT-5 will be offered to all users, from those who use them for free through ChatGPT to those who work in companies that pay for large institutions plans. There, of course, some differences between how they appear based on whether you pay for it. This is a collapse:

• Free Users: You will get access to GPT-5 until the cover cover, after which you will have the lighter GPT-MINI.
• In addition to users: Similar to free users, but with the limits of the highest use.
• Professional users: Unlimited access to GPT-5 and access to the most powerful GPT-5 Pro model.
• Foundation users/EDU/Team: The GPT-5 will be the default model, although it may be next week before it is presented to everyone.

GPT-5 itself is actually some different models. There is a fast llm but somewhat direct and more powerful Thinking model To deal with the most complex questions. The guidance program determines any model that can deal with the claim.

Even coding skills are faster

The Openai is especially highlighted the skills and speed that the new model can write code. This is not just a job for programmers. The ability of the model to write a program that makes it easy for it to solve a problem that you present to it by creating the right tool.

Yann Dubois, post -training progress in Openai, showed the ability to codish the model by asking to create an application for French learning. Within minutes, she encoded a full web app with audio and work game functions. Dubois has already requested the creation of two different applications, working on the same router through the model twice. The symbol that you write, which you write, allows you to try several times, and choose the result that you like better-or make notes to make changes until you get it properly.

“Beauty is that you can repeat Super quickly with the GPT-5 to make the changes you want,” said Dubuis. “GPT-5 really opens a fully new world of coding.”

Read more: Never use Chatgpt for these 11 things

New safety features

After announcing some steps this week to improve how its tools Dealing with sensitive mental health problemsOpenai said that GPT-5 has some adjustments to make things more safe.

Alex Biotel, Safety Research Commander, said that the new model has improved training to avoid deceptive or inaccurate information, which will also improve the user experience.

It will also respond differently if you ask a claim that may be dangerous. Biotel said the previous models will refuse to answer a possible harmful question, but the GPT-5 will instead try to provide the best safe answer. This can help when the question is innocent (such as a science student who raises a chemical question) but he looks more evil (like someone trying to create a weapon). “The model is trying to give an answer as much as possible, but within the restrictions of a sense of safety,” Biotel said.

But is this really the way to AGI?

Altman told reporters that the model is an “important step along the way to AGI”, or artificial general intelligence, a term that often refers to smart and capable models like humans. But Altman also said that it is certainly not yet. One of the major reasons is that he is still not learning continuously while publishing it.

Openai’s declared goal is to try to develop AGI (although Altman said it is not a fan of the term), and he has won a competition. Mark Zuckerberg, CEO of Meta, is hiring senior artificial intelligence scientists with the aim of creating “Superintelligence“

Whether the big language models are the way there, no one knows now. Three quarters of artificial intelligence experts Earlier this year survey They said they have doubts that LLMS would expand the scope of something of this level of intelligence.