OpenAI Unveils GPT-4.5 with Enhanced 'EQ' – Key Details

OpenAI, an AI company supported by Microsoft, has released its newest chat model, GPT-4.5, which it claims is the largest and most sophisticated so far. The company stated that the new model enhances the quality of conversations, user intent understanding, and “EQ” (emotional quotient) abilities. As a research preview, GPT-4.5 is accessible to ChatGPT Pro users and developers around the globe.  

GPT-4.5: What has changed?  

According to OpenAI, GPT-4.5 expands its focus to include reasoning beyond STEM, achieving broader capabilities. The model was trained using a combination of new supervision techniques, as well as the tried and tested methods of supervision through fine-tuning (SFT) and training with reinforcement from humans (RLHF), just like GPT-4o did.  

The company emphasized that models enhanced with steering and nuance understanding perform better, which is why GPT-4.5 was trained on the data of these smaller models. This results in more natural interactions and better alignment with user intent. The model also has a larger scope of information.

Unlike OpenAI’s reasoning models, GPT-4.5 “doesn’t think before it responds.” Instead, it specializes in reading deep subtleties and having emotional responses which make it useful for writing, designing, programming, and other real-world tasks. OpenAI mentioned that they are still trying to understand what all the model can do and assume it will have some surprising uses.

Regarding safety, OpenAI said that GPT-4.5 internal evaluations did not increase risks compared to other existing models. In addition, the model is expected to lessen hallucinations and give more accurate answers to many subjects.

GPT-4.5: Availability 

GPT-4.5 is now available on the web, mobile, and desktop for ChatGPT Pro users through the model picker. Plus and Team users will receive it next week and Enterprise and Edu users will receive it the week after. 

Users can still use features such as Search, Canvas, and file upload features with GPT-4.5. However, at this time, ChatGPT doesn’t support multimodal features like Voice Mode, video, or even screen sharing.