Open AI introduced ChatGPT 4.5 in late February 2025 calling it their “largest and best model for chat” so far scaling the feature of unsupervised learning to the next level. Enhancing its ability to recognize patterns, draw connections, and generate creative insights without reasoning, ChatGPT 4.5 is now available to use for free users and has shown a broader knowledge base, ability to understand user’s prompts better with a higher level of EQ.
In a video podcast shared on Open AI’s official YouTube channel featuring founder Sam Altman, Alex Paino, Amin Tootoochian and Daniel Seisam working with algorithms system architecture, they discussed about the pre-training process of ChatGPT 4.5.
What’s the new ChatGPT update?
The team has advanced AI capabilities by scaling two complementary paradigms, unsupervised learning and reasoning, the two apex intelligences. While unsupervised learning increases the world model accuracy, an update from ChatGPT 3, 3.5 and 4o, Scaling reasoning, on the other hand, teaches models to think and produce a chain of thought before they respond, allowing them to tackle complex STEM or logic problems. Models like OpenAI o1 and OpenAI o3‑mini advance this paradigm.
Achieving more than 60% accuracy in its responses, the pre-training of ChatGPT 4.5 integrated machine learning and system teams. This mammoth task was focused on responding to unresolved issues rather than prioritising an earlier release. Essentialy, the ultimate goal of the team was to achieve a ten times smarter version of 4o.
What’s next for ChatGPT?
The team has scaled up from 10,000 to 100,000 GPUs after the Ghibli style image trend took the internet by storm and overloaded the Open AI servers. For GPT‑4.5, we developed new, scalable techniques that enable training larger and more powerful models with data derived from smaller models. These techniques improve GPT‑4.5’s steerability, understanding of nuance, and natural conversation.
As per founders, GPT‑4.5 doesn’t think before it responds, which makes its strengths particularly different from reasoning models like OpenAI o1. Compared to older versions, GPT‑4.5 is a more general-purpose and innately smarter model.
