ChatGPT 5 is anticipated to bring advanced reasoning abilities.
Since its launch, OpenAI’s ChatGPT made waves in generative AI, with the company keeping up the pace by releasing enhanced GPT models over the past two years. Competitors like Google and Meta are advancing, but the race in generative AI will persist until we reach artificial general intelligence (AGI).
OpenAI, a leading AI startup, is reportedly working on a new large language model called Orion, also known as GPT-5, which is unrelated to Meta’s Oryon AR smart glasses.
Several reports suggested that OpenAI’s next AI model might launch as soon as December 2024. However, a company spokesperson recently refuted this in a statement to The Verge, which could be linked to the recent departures of top executives, including former CTO Mira Murati.
According to The Verge, Orion might not see a public release like GPT-4. Instead, it’s expected to be a model initially available to OpenAI partners, such as Microsoft, to support services like Copilot.
Since GPT-4’s release in March 2024, OpenAI has rolled out various updated versions, like GPT-4o and OpenAI o1. We may not see an official update on the secretive Orion—likely to be called GPT-5—until early 2025.
Orion may be significantly closer to achieving artificial general intelligence (AGI).
GPT-5 could become the most advanced large language model yet, with outstanding reasoning abilities. Reports suggest OpenAI might complete its pre-training by December 2024, with additional months planned for fine-tuning before the official launch.
However, multiple reports indicate that it will not reach the level of artificial general intelligence (AGI) but will serve as a step toward achieving the pinnacle of artificial intelligence. GPT-5 is expected to surpass GPT-4o in nearly every aspect, which can be categorized into three key areas: a larger context window, an expanded knowledge base, and enhanced reasoning abilities.
OpenAI’s former CTO, Murati, recently stated that “systems like GPT-4 resemble the intelligence of a smart high school student.” She suggested that in the coming years, we could see AI models with PhD-level intelligence for specific tasks, emphasizing the rapid pace of advancements. OpenAI co-founder Altman has also noted that the next model will represent a “significant leap forward,” although he acknowledged that there is still considerable work to be done.
OpenAI’s GPT-5 is expected to be a multimodal AI model, capable of processing text, audio, and image inputs while also generating text, audio, images, or combinations of these formats. Like Llama 3.2 and Gemini, GPT-5 may be available in various sizes, each tailored for specific applications.
The effectiveness of an AI model is often assessed based on the number of parameters it is trained on, and the upcoming GPT-5 is expected to be one of the largest large language models, reportedly featuring over 1.5 trillion parameters.
Achieving human-level intelligence is nearly assured.
At the recent Build with AI summit in India, Meta’s chief AI scientist Yann LeCun remarked that even the most advanced AI models are not as intelligent as a four-year-old. He noted that many AI companies are striving to develop models with human-level intelligence, and GPT-5 could be among them.
Safety remains a primary concern regarding AI models. Many of these models implement various measures to address this, such as watermarking AI-generated content, refraining from using personal data for further training, and preventing bias or malicious behavior. In a recent report, a mother alleged that AI caused her son’s death. OpenAI is expected to tackle most of these issues rigorously in GPT-5.