2 min Applications

GPT-5 aims to end AI model overgrowth at OpenAI

GPT-5 aims to end AI model overgrowth at OpenAI

There has been a proliferation of AI models within ChatGPT. Multiple AI models, each excelling at different tasks, need to be consolidated. That is OpenAI’s plan with GPT-5, whenever it is released.

GPT-5 has been mentioned occasionally in recent years by OpenAI staff. However, no new number has been added to the Generative Pretrained Transformer on which ChatGPT runs since April 2023. Instead, variants of GPT-4 and reasoning o-models have been reviewed. According to VP of Research at OpenAI Jerry Tworek, its ultimate successor is well on its way.

All in one

OpenAI CEO Sam Altman has previously expressed his frustration with the list of AI models that ChatGPT and OpenAI API users have to sift through. While GPT-4o excels at everyday tasks, o3 is intended for in-depth answers, but there are also Deep Research and search engine buttons. In fact, it is up to the user to determine which query best suits each individual model.

With GPT-5, this “model switching” should be a thing of the past, according to Tworek. Instead, the next OpenAI model will be multimodal and may switch seamlessly between what were previously separate AI models. Think of the AI image generator based on 4o. Currently, this cannot be consulted when a conversation with o3 is taking place. With GPT-5, this should work more intuitively.

UI issues

We haven’t yet mentioned the domain-specific options within ChatGPT, which also need to be captured by GPT-5 for OpenAI to consolidate into a single model. For example, OpenAI is currently working on Codex as an AI coding assistant and agent, while also offering the not-so-popular Operator tool for controlling a PC.

The fact is that OpenAI users do not only use ChatGPT. Frustrations about the slow and glitchy UI have given rise to alternative chat apps with a choice between OpenAI’s offering and that of its competitors.

In addition, the API is extremely popular; it remains to be seen how GPT-5 will play into that. Currently, the model choice allows API users to save costs with, for example, o3-mini or GPT-4o mini.

Read also: Anthropic raises bar for GPT-5 with ‘Artifacts’ feature