In May this year, OpenAI unveiled “GPT-Next” for the first time at an event in Paris. This reveal surprised many. GPT-Next is supposed to be the successor to GPT-4. What about GPT-5? There is a completely different story going on on this front, with Project Strawberry being announced, and a new AI model called Orion being worked on. But that is a story for another time. Regarding GPT-Next, there have already been promises of better multimodal capabilities and smarter training models that can significantly improve the functionality of AI. Now, at the KDDI Summit 2024, Tadao Nagasaki, CEO of OpenAI Japan, has revealed that the next AI model called GPT-Next is coming soon, and it will be 100 times better than GPT-4.

According to Nagasaki, the GPT-Next model will use a smaller version of OpenAI’s secretive Project Strawberry, also previously known as Project Q* or Q-Star. Project Strawberry is believed to be OpenAI’s attempt to create artificial general intelligence. Additionally, GPT-Next is said to have improved performance due to improvements in the model’s architecture rather than relying heavily on huge computing resources.

During the event, Nagasaki revealed that by the end of August, ChatGPT had surpassed 200 million active users, making it the fastest software ever to reach this milestone. He emphasized that no other software has made AI accessible to users in such a user-friendly way. The CEO of OpenAI Japan also talked about the future of the GPT series, making a comparison between GPT-3 and GPT-4. He pointed out that unlike traditional software, AI is moving fast, and OpenAI is committed to accelerating the integration of AI in the world.

Its improvements in AI are truly exponential. The next big AI model to be launched by OpenAI is 100 times better than its current flagship model, GPT-4. Even bigger improvements are expected with Orion and GPT-5 Strawberry. For reference, in May, OpenAI CTO Mira Murati announced a new model called GPT 4o, which was already “much faster” than GPT-4. “GPT4o is the future of interaction between humans and machines,” Murati said. “It improves on its (GPT 4) capabilities such as text, vision and audio,” she adds. GPT 4o is 2 times faster, 50 percent cheaper and has 5 times higher rate limits than GPT 4 Turbo. Now imagine a 100 times improvement.

Possibly talking about GPT-Next, Murati said that the model could in some way be smarter than humans. Speaking at his alma mater, Dartmouth Engineering, he talked about the future of AI and the next version of ChatGPT, which he described as having PhD-level intelligence for specific tasks. According to Murati, GPT-3 demonstrated intelligence similar to that of a child, GPT-4 had the cognitive abilities of a high school student, and the upcoming model would be even more advanced. He mentioned that with the new version, in many conversations, users may find that the next-generation chatbot appears to be more intelligent than themselves.