GPT-5 could be here this month — here's five features we're hoping to see

Tom's Guide

17 hours ago

For fans of ChatGPT, a big moment could be just around the corner. OpenAI has been dangling something in front of us for a while — the launch of GPT-5.
This would be a major update for ChatGPT, bringing with it some huge changes to the functionality of the famous AI chatbot. It has been delayed, hinted at, and hyped up for months, but if rumors are correct, we should see it in July.
However, rumors only tell us so much. Sam Altman, CEO of OpenAI, has only said 'summer' as the expected release date. While others have hinted at a July launch, it could still be a few more months, especially with OpenAI dealing with lawsuits and staff getting poached by Meta AI.
Whether it's a July launch or later, we here at Tom's Guide have high hopes for this latest update. But what will it bring about? These are the features we're hoping for most.
This stands out as both the feature we're hoping for most, as well as the one that OpenAI is likely to put the biggest focus on.
AI companies are pouring time and energy into improving their models' ability to reason and plan. In other words, how well can ChatGPT analyze holistically and bring an answer together?
This, to some people, will be concerning; it is an important step for AI towards thinking, or at least as much of a step as an AI can take. Instead of following algorithmic steps, it could take a more human approach to things.
Get instant access to breaking news, the hottest reviews, great deals and helpful tips.
Developments in this area could also see an end, or decrease, to one of AI's biggest issues of hallucinations. Limiting how much AI makes mistakes is crucial. Giving it more time to think and process its answers helps to limit this.
On top of all of this, better reasoning could see ChatGPT launch forward in its coding ability and understanding of more complex prompts.
Along with the above changes, GPT-5 is likely to solve some of ChatGPT's personality traits. In other words, making it interact in a more natural, human way.
Some other AI chatbots, such as Claude, have started introducing the ability to give specific personality traits or response styles in conversations. While it is unclear if ChatGPT will have something like this, it will likely at least be more natural in its conversation style.
This will especially appeal to anyone who likes to have back-and-forth conversations on the chatbot, or equally, if they like using it to practice real-life conversations they might soon have.
Currently, ChatGPT has a limit on how much information you can feed it. This can be in terms of the size of documents, transcript lengths or number of pages of a document. In other words, it can only process so much information. Right now that ranges, depending on the version you're using, but can go up to around 128,000 tokens.
That sounds like a lot, and in a way it is. However, with the improvements that GPT-5 could bring it may jump drastically. OpenAI has increased the context window with each release of a new model, and GPT-5 seems to be its biggest gain yet.
This could be especially useful for long conversations where ChatGPT is fed multiple files and pieces of information to piece together. For example, giving the chatbot a year of expense files to look through and process the information.
While this jump hasn't been promised or alluded to by OpenAI, it is highly likely.
Naturally, with all of these improvements, whether it is to the context understanding, increase in training or an improved memory, ChatGPT will also see improvements in its ability to create images and videos.
While Sora isn't necessarily the best in business for video generation, this update could bring it up there to compete with Gemini's Veo 3.
Image-wise, ChatGPT is already in a great position, but could see further improvements. Even something as simple as being able to process context better, or take in more details mentioned in your prompts.
Currently, ChatGPT offers 'GPTs'. These are customized versions of the chatbot, designed for different tasks. This could be grammar checkers, an expert on Space or some other niche subject.
However, this can also be used to offer access to 3rd party tools. You can use ChatGPT to make reminders on Any.do for example. While there has been no announcement of improvements in this area, I'm hoping to see OpenAI make it easier to pair ChatGPT with other apps.