OpenAI has unveiled GPT-5 – a more powerful model for coding, writing, and complex tasks

OpenAI has unveiled its new GPT-5 artificial intelligence language model, which offers significantly improved capabilities for programming, creative writing, and complex logical thinking.

The presentation took place during an online event on August 7.

GPT-5 can:

• respond to requests faster;

• make up facts less often;

• better justify its answers;

• independently determine how much time to spend on processing complex requests;

• choose one of four interaction styles: “Cynic,” “Robot,” “Listener,” and ‘Nerd’ (currently in “research preview” format).

According to the company’s CEO Sam Altman, this is a “major update” that for the first time creates the feeling that you are communicating with an expert on any topic, Bloomberg reports. 

OpenAI CEO Sam Altman at the presentation of the new GPT-5 model. Screenshot

GPT-5 has been available to ChatGPT users and most subscribers since Thursday, and business and education customers will be able to use it fully starting next week (August 11).

Progress in programming

OpenAI has paid particular attention to the use of GPT-5 in programming — both when creating new code and for debugging it. The company believes that the idea of “software on demand” will become the main trend of the new GPT era.

The new GPT-5 model demonstrates programming skills during a presentation. Screenshot

During the demonstration, GPT-5 created an interactive application for learning French with flashcards, a quiz, and a game — all in just a few minutes.

The model is already being tested by companies such as Anysphere (developer of the Cursor programmer assistant) and Swedish startup Lovable. According to the latter’s CEO, GPT-5 is more effective than other models at creating complex applications and generating code that is easy to maintain.

It imitates, rather than replicates

The company claims that the new model has been trained to be more honest, give more accurate answers, and generally behave more “humanly.”

However, some experts who shared their thoughts on the new model with the BBC question how new it really is. They point out that these systems still aren’t profitable and still imitate, rather than replicate, human thinking.

In their opinion, there’s a chance that without constant hype, interest in the industry will drop.

BBC journalist Marc Cieslak, who got exclusive access to GPT-5 before it officially came out, says, “It’s more of an evolution than a revolution in tech.”

GPT-5 thinks “deeper,” but how it works is pretty much the same: users ask questions or give tasks in a text box.

Read more: Google launched Guided Learning tool in Gemini — a response to ChatGPT’s Study Mode