GPT-5 Launches This Summer: Explore the Exciting New Features Awaiting You!

The release of GPT-5 is fast approaching. In a podcast released two weeks ago, Sam Altman mentioned that OpenAI’s next significant model is set to debut this summer. What can we expect from it? Let’s delve into the details.

What’s New in GPT-5?

True to form, OpenAI is keeping the specifics about the new features of GPT-5 under wraps. However, as development progresses and announcements are made, certain features are becoming apparent. The following characteristics are expected in the next model:

• Chain-of-thought reasoning: In February, Sam Altman was quite explicit: GPT-4.5 is the last model without chain-of-thought reasoning. All future models, including GPT-5, will inherently incorporate step-by-step reasoning. “One of the things that really excited people about the initial GPT models was the ability to achieve better performance by instructing the model to ‘think step-by-step’… The reasoning models will take this even further,” he explained in the podcast.

Stunned detectorists uncover 1,400-year-old Anglo-Saxon gold treasure—experts say missing eye may hold secret meaning

Meet Finn: The Astonishing Cat as Big as a 9-Year-Old Child!

• Enhanced memory: Following recent updates to memory capabilities, GPT-5 is expected to have improved long-term memory for ongoing conversations. According to Sam Altman, this will save users time as the AI will already know their preferences: “The computer now knows enough about me and my life to anticipate what I want it to do, sometimes in ways I wouldn’t have thought of myself.”
• Improved multimodal capabilities: Since the introduction of GPT-4o, OpenAI has emphasized multimodality as a key feature, highlighted by the introduction of advanced vocal capabilities. Considering the latest innovations, such as the new image generator, multimodality is expected to be a central element in the next iteration. “You might want something that is much more aware of its surroundings, that fits much more into your life context. You might want to interact with it in ways other than just typing and looking at a screen,” Sam Altman noted.
• A more agentic orientation: Another direction that OpenAI has taken in recent months, which is likely to continue with GPT-5, is the agentic dimension of AI, meaning its ability to interact with software interfaces or digital environments. This approach began with Operator thanks to the o3 model and represented a “significant step forward” for Sam Altman. It is without doubt that GPT-5 will continue in this direction, especially since it will incorporate the capabilities of o3.

GPT-5: A Less Radical Transition

The transition from GPT-3.5 to GPT-4 marked a significant leap forward for OpenAI. However, models have been updated very regularly since then. Thus, the shift to GPT-5 is expected to be seen as less revolutionary. “Previously, things were much clearer. We would train a model and release it, then we would train a new large model and release it… Today, systems have become much more complex and we can continuously refine them,” explains Sam Altman.

We could continue to iterate on 4.5 and, at some point, just decide to call it five. This is something we are currently considering. Suppose we launch GPT-5 and then improve it, over and over. Do we keep calling all of this GPT-5, as we did with GPT-4o? Or should we call it 5.1, 5.2, 5.3, so that users can know when a version has changed, wonders the CEO of OpenAI.