According to a new report from Business Insider, OpenAI is expected to release the new LLM for ChatGPT very soon, an evolution of the previous GPT-5, an improved version of the AI language model that powers ChatGPT, in mid-2024, likely over the summer. Two anonymous sources close to the company revealed that some enterprise customers recently received GPT-5 demos and talked about related improvements to ChatGPT.
One of the two sources described the model as “really good” and “substantially better” than what we saw with GPT-4, and in that case OpenAI would show the new LLM using use cases and data specific to the company in question. The anonymous CEO of this company also hinted at other capabilities that have not yet been released, such as the ability to launch artificial intelligence agents developed by OpenAI to perform tasks automatically. Of course, the name GPT-5 is just speculation, as it could also be an intermediate incremental update, such as GPT-4.5, for all we know.
Recall that OpenAI launched GPT-4 in March 2023 as an update to its main predecessor, GPT-3, which came out in 2020, but arrived at GPT-3.5 in late 2022). Last November, OpenAI released GPT-4 Turbo, which greatly reduced the inference costs of OpenAI’s LLM, however, it was not without its critics. We can say that it has often been referred to as “lazy”, as sometimes the model refuses to respond to prompts or complete coding projects as required. OpenAI has tried several times to solve the problem, but apparently without ever fully succeeding.
The next and expected version in a few months should be a multimodal language model that can accept text or encoded visual input, and like GPT-4, it will be a prediction model of the next token, i.e. it will provide its best estimate of the next most likely token (a word fragment) in a sequence, allowing tasks such as completing a sentence, writing code, and much more.
Recall that LLMs are trained on massive datasets pulled from the internet and licensed by media companies (or at least that’s how it should be), allowing them to respond to user prompts in a human-like way. However, the quality of the information provided by the model can vary depending on the training data used, and also based on the model’s tendency to invent information, known in jargon as “hallucinations”, something that practically everyone has experienced while using a chatbot. If GPT-5 can improve its capabilities and reduce errors, it will represent a significant advance for the company.
According to the source, OpenAI is still training GPT-5 and after that the model will undergo internal security testing to address any issues before public release. The release date may be delayed depending on the duration of these processes, so summer 2024 is purely indicative.
