OpenAI, a name that has become synonymous with the AI revolution after the release of their flagship product ChatGPT has just announced GPT-4.5.
The new model is set to be the ‘largest and best model for chat yet.’
But what is GPT-4.5, how is it different from previous models and how can you access it? We’ve got you covered with this comprehensive guide to OpenAI’s latest innovation.
What is GPT-4.5
GPT-4.5 is the most advanced model of OpenAI’s large language model. The AI giant states the advancement will be a huge step forward in scaling up pre-training and post-training.
Initial testers report that GPT-3.5 feels more like natural conversation than using its predecessors. This is likely due to its wider knowledge and improved abilities in understanding user intent.
Read: OpenAI's Deep Research Agent Will Conduct Novel Scientific Research
Crucially OpenAI are making waves in addressing the major issue of AI and expect 4.5 to hallucinate less.
The two key changes that advance GPT-4.5 are two complementary paradigms: unsupervised learning and scaling reasoning.
What is unsupervised learning?
Unsupervised learning is a type of machine learning where AI models are trained on a lot of unlabeled data. The AI then learns to recognize patterns and structures within the data without explicit guidance from human input.
Through processing the massive amounts of data the AI is able to build an internal representation of how the world works, this includes relationship between concepts, ideas and object- this is known as an internal world model.
By building a better internal world model the AI model develops a computational form of ‘intuition’ which helps it to make more accurate predictions, understand nuances and allow for more flexible responses.
What is scaling reasoning?
Scaling reasoning is the practice of increasing the size and complexity of AI models, along with the amount of training data, to enhance reasoning abilities.
Larger models can typically learn more intricate patterns and relationship which lets them handle more complex tasks. OpenAI claims GPT-4.5 can apply a chain of thought reasoning before it responds which will allow it to tackle complex STEM or logic problems.
How To Use GPT-4.5
GPT-4.5 is currently available as a research preview. This means the model is still being refined.
ChatGPT Pro users can access GPT-4.5 by selection the option in the model picker on mobile, web or desktop chat windows.
The second phase of the rollout will include Plus and Team members in the next week, followed by Enterprise and Edu users next.
Comments ( 0 )