AI BriefingOpenAI Unveils GPT-4.5
Key point
OpenAI has released a research preview of GPT-4.5, which scales up unsupervised learning to strengthen knowledge and intuition.
Details
OpenAI has released a research preview of GPT-4.5, its most powerful GPT model. This model focuses on significantly scaling up unsupervised learning to strengthen pattern recognition, the ability to make connections, and creative insight.
AI advances along two axes: unsupervised learning and reasoning. GPT-4.5 demonstrates progress in unsupervised learning by scaling up compute resources and data massively to improve the accuracy and intuition of the world model. This is a different approach from reasoning-scaling models like OpenAI o1, which use a 'chain of thought' to solve complex logical problems.
Trained on a Microsoft Azure AI supercomputer, GPT-4.5 has the following characteristics:
- A broader knowledge base and improved ability to grasp user intent
- High 'EQ (emotional intelligence)' that supports natural interaction
- Reduced hallucination and higher reliability across a variety of topics
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.