GPT-4.5 OpenAI’s Largest Model Steps Into the Spotlight

OpenAI is launching GPT-4. 5 today, the latest and biggest of its language models. And while it’s being presented as the company’s “largest and most knowledgeable model yet,” OpenAI warns that it’s not a frontier model and might underperform when compared to o1 or o3-mini.

GPT-4. 5 offers improved writing skills, upgraded knowledge of the world, and a “refined personality” over past models. OpenAI notes that GPT-4. 5 is more than an improvement on GPT-4’s computation efficiency (+10x), but it’s not a frontier model due to not enough new capabilities.

GPT-4’s evolution. Reportedly, 5 used OpenAI’s o1 reasoning model (codenamed Strawberry) to generate synthetic training data. The training incorporated new supervision techniques alongside more traditional supervised fine-tuning and reinforcement learning from human feedback, in much the same way as the GPT-4o process.

One significant upgrade is that GPT-4. 5 hallucinate considerably less than GPT-4o and marginally less than the o1 model. According to reports, access to GPT-4. Like Operator, ChatGPT 5 will on launch be available only to ChatGPT Pro subscribers.

This release is consistent with previous reporting that OpenAI intended to make GPT-4 available. five by the end of February, with GPT-5 potentially following as soon as late May. OpenAI CEO Sam Altman has said that, at its core, GPT-5 is “a system that integrates a lot of our technology,” including the new o3 reasoning model that the company previewed in December. Although o3-mini was released earlier last month the full o3 model will only become available as a part of the upcoming system of GPT-5, all of this aligns to OpenAI’s aim of producing a more capable model that might resemble artificial general intelligence someday.

Written By :

Valentina Morelli