OpenAI unveils GPT-5 with enhanced coding capabilities and safety

All users, including free tier, will have access to GPT-5 model
An undated image. — OpenAI
An undated image. — OpenAI

OpenAI finally revealed its newest flagship model, GPT-5, on Thursday, after months of waiting.

In comparison to its predecessor, GPT-4, the new model is expected to provide improved speed, accuracy, and performance.

OpenAI CEO Sam Altman said, "I tried going back to GPT-4, and it was quite miserable." Altman went on to stress that GPT-5 is "significantly better in obvious and subtle ways."

According to the company, ChatGPT and other apps that make use of OpenAI's technology will be powered by GPT-5, which will also power the chatbots, agents, and search tools.

OpenAI's GPT-5 model tier structure

All users, including those on the free tier, will have access to the GPT-5 model, although there will be usage restrictions. The following will be the tier structure:

  • Free users will be forced to switch to the lighter GPT-5-mini model after using GPT-5 for a predetermined amount of time. 
  • Plus, users will have access comparable to that of GPT-5, with greater usage restrictions.
  • Pro users will have unrestricted access to GPT-5 and a more robust GPT-5 Pro model. 

The rollout may take a week, but GPT-5 will be the default model for Enterprise, EDU, and Team users.

What to expect from OpenAI's GPT-5 model?

Additionally, the model's enhanced coding abilities have been emphasised by OpenAI, which has shown that it can swiftly produce unique applications without the need for coding.

OpenAI Post-Training Lead Yann Dubois demonstrated the model's coding skills by asking it to create a French language learning app.

The model produced a web application with functional game features and sound in a matter of minutes. "The beauty is that you can iterate super quickly with GPT-5 to make the changes that you want," Dubois said.

Moreover, the updated model has better training to steer clear of false or misleading data.

GPT-5 will react differently to potentially dangerous prompts, giving the best safe response rather than not responding at all, according to Safety Research Lead Alex Beutel.

 He said, "The model tries to give as helpful of an answer as possible but within the constraints of feeling safe."

Furthermore, all users will now have access to the Advanced Voice mode, albeit with more usage restrictions.

Users will be able to alter the colour of their chats, though some features will only be available to those who have paid.

Additionally, ChatGPT will be able to link to Gmail and Google Calendar accounts, allowing users to ask questions about their schedule and get recommendations.

According to Altman, GPT-5 is a "significant step along the path to AGI" (artificial general intelligence). He did stress, though, that the model is not yet able to learn continuously while in use. OpenAI is working towards its stated goal of developing artificial general intelligence (AGI).

A related development is that Meta's CEO, Mark Zuckerberg, has been hiring leading AI researchers in an effort to develop "superintelligence". There is still disagreement over whether large language models are necessary to achieve AGI.

An important turning point in the advancement of artificial intelligence has been reached with the launch of GPT-5.

The model promises to offer a more advanced and user-friendly experience with its improved coding capabilities, safety features, and extra functionality.