OpenAI’s next major model may be approaching, but the clearest picture so far is still a carefully hedged one. A Business Insider report says GPT-5, described as an improved version of the AI language model behind ChatGPT, is expected sometime in mid-2024 and likely during the summer.
The report also suggests that OpenAI has already shown GPT-5 and related ChatGPT enhancements to some enterprise customers. That does not make a launch date official. It does, however, give the strongest reported signal yet that the next major ChatGPT upgrade is far enough along to be demonstrated outside the company.
What the report says about GPT-5
According to the Business Insider report, two anonymous sources familiar with OpenAI said some enterprise customers have recently received demos of GPT-5. One CEO who saw a version of the model described it as “really good” and “materially better.”
The same CEO said OpenAI demonstrated the model with use cases and data unique to his company. That detail matters because it points to the practical question many ChatGPT users care about most: whether a new model can be more useful in specific work settings, not just more impressive in general conversation.
The report also said the CEO hinted at other unreleased capabilities, including the ability to launch AI agents being developed by OpenAI to perform tasks automatically. The source article does not provide further details on how those agents would work, when they would appear, or whether they would arrive with GPT-5.
OpenAI has not confirmed the release date
OpenAI representatives were asked about GPT-5’s release date and the Business Insider report. They had no particular comment, but they pointed to a recent appearance by Sam Altman on the Lex Fridman podcast.
Lex Fridman(01:06:13) So when is GPT-5 coming out again?
Sam Altman(01:06:15) I don’t know. That’s the honest answer.
Lex Fridman(01:06:18) Oh, that’s the honest answer. Blink twice if it’s this year.
Sam Altman(01:06:30) We will release an amazing new model this year. I don’t know what we’ll call it.
Altman’s answer leaves the branding unresolved. The company may release something that looks like a GPT-5-level model, or it may choose another name. The source article notes that it could be considered a major upgrade to GPT-4 Turbo, or perhaps an incremental update like GPT-4.5.
That uncertainty is important. The market may be waiting for the label “GPT-5,” but OpenAI’s public comments leave room for a different naming decision. What matters for users is whether the model improves the capabilities inside ChatGPT and related products.
Why the next model matters
GPT-4 launched in March 2023 and created intense attention because of its ability to imitate human communication and composition. Since then, the question of when GPT-5 might arrive has followed OpenAI CEO Sam Altman through interviews and media appearances around the world.
The source article describes GPT-5, or whatever it is ultimately called, as expected to be a multimodal large language model. Like GPT-4, it is expected to accept text or encoded visual input, known as a “prompt.”
It is also expected to remain a next-token prediction model. In plain terms, that means the system generates output by estimating the most likely next token, a fragment of a word, in a sequence. That mechanism can support tasks such as completing a sentence or writing code, and when configured for conversation, it can power chatbot applications like ChatGPT.
OpenAI’s model history gives the rumored release context. GPT-3 emerged in 2020. GPT-3.5 arrived in late 2022. GPT-4 followed in March 2023, and GPT-4 Turbo came last November.
The safety process could affect timing
The Business Insider report says OpenAI is still training GPT-5. After training is complete, the model is expected to go through internal safety testing and additional “red teaming” before any public release.
That process is not a minor final step. The source article says the release date could be delayed depending on how long safety testing takes. It also notes that the report’s sources could be mistaken, and that GPT-5 could launch later for reasons aside from testing.
Several factors appear central to whether GPT-5 would represent a notable advancement:
- Whether it improves generalization, meaning its ability to perform novel tasks.
- Whether it reduces what the industry commonly calls “hallucinations.”
- Whether it addresses user concerns that followed GPT-4 Turbo, including accusations of “laziness” where the model sometimes refuses to answer prompts or complete coding projects as requested.
LLMs like OpenAI’s models are trained on massive datasets scraped from the Internet and licensed from media companies. The quality of their answers can vary depending on that training data and on the model’s tendency to confabulate information.
For that reason, the reported promise of a “materially better” ChatGPT upgrade should be read alongside the remaining unknowns. The report points to progress, demos, and a possible mid-2024 window. OpenAI’s own public stance remains more cautious: an amazing new model is coming this year, but the name and timing are not settled.