The Evolution of LLMs: A comparison of GPT-5 vs. GPT-4

"People have these unrealistic expectations that GPT-5 is going to be doing back flips in the background in my bedroom while it also writes all my code for me and talks on the phone with my mom or something like that," said Logan Kilpatrick, Head of DevRel at OpenAI, on an episode of Lenny's Podcast. "I'm like, 'That's not the case.' It's just going to be this very effective tool, very similar to GPT-4, and it's also going to become very normal very quickly."

So, what can be anticipated when comparing GPT-5 to GPT-4?

Advanced reasoning: In a recent podcast with Bill Gates, Altman emphasized the need to enhance the reasoning abilities of GPT-4, as it currently has limited capacity in this area. However, GPT-5 is set to address this issue by having more advanced reasoning capabilities. This means that the model will be able to analyze complex data sets and solve more sophisticated problems.

Better accuracy and speed: In terms of accuracy, GPT-4o has seen significant improvements in reasoning, with accuracy rates on par with GPT-4 Turbo. It can answer general knowledge questions accurately with an impressive 87.2% accuracy. On the other hand, GPT-5 is set to push the boundaries even further with its ability to process up to 50,000 words at a time, which is twice as many as GPT-4 can handle. This faster processing speed will undoubtedly enhance the overall performance and efficiency of the model.

Extended multimodality: One of the most exciting features of GPT-5 is its multimodality capabilities. It is expected to build upon the existing features of GPT-4 and introduce new functionalities like video integration. This will significantly change how we communicate with AI, allowing for more natural and flexible forms of interaction.

Increased reliability: Sam Altman emphasized that one of the key improvements in GPT-5 will be increased reliability. He stated that while GPT-4 might provide a pretty good response if asked the same question 10,000 times, it doesn't always know which response is the best. With GPT-5, the goal is to consistently provide the best response out of those 10,000, thus significantly improving reliability.

Overall, the advancements in GPT-5 compared to GPT-4 are promising. With improved reasoning abilities, higher accuracy rates, faster processing speeds, and enhanced multimodality features, GPT-5 is set to further advance the field of artificial intelligence. It will open up new possibilities for AI applications and pave the way for more intelligent and efficient AI systems in the future.

At this time, OpenAI has not announced an official release date for GPT-5. However, the launch is anticipated to occur between May 2024 and December 2024. As with previous models, there may be delays or changes in the timeline, so it is best to keep an eye out for any updates from the company.

In terms of cost, GPT-5 is expected to follow a similar payment structure as its predecessors. While ChatGPT has a free version and subscription plans (Plus, Team and Enterprise), the API-based services for GPT-5 are expected to once again be based on the usage amount. The pricing structure includes costs per input token for the text you send and per output token for the text generated by the model. As GPT-5 can also be expected to be the most capable LLM of OpenAI, pricing will likely not be below OpenAI’s currently best performing model GPT-4o.

I believe that OpenAI's main focus is to enhance the reliability and reasoning capabilities of GPT-4. I find the Voice Mode of GPT-4 quite interesting and anticipate further improvements in that area. In my view, we can expect a transition similar to the one we experienced between GPT3.5 to GPT-4.

-Thanh Long Phan, Machine Learning Scientist

Our team is eagerly anticipating the remarkable advancements in AI, particularly the evolution of language models. The recent progress in this field is truly inspiring, and we are glad to be in the center of developing state-of-the-art applications on top of large language models (LLMs). Within our organization, we are dedicated to using these advancements to create customized LLM solutions for our clients in different industries.

Discover our services that use the latest AI and language modeling to transform your organization with significant changes.

At dida, we are specialized in developing custom AI solution for mid to large-size enterprises.

If you are interested in custom solutions related to the models of OpenAI for your company, reach out to us. We're happy to discuss your custom projects.