OpenAI has released its latest model, GPT-4.5, for ChatGPT Plus subscribers, promising enhanced emotional intelligence and nuanced understanding of user input. But how does it compare to the current standard model, GPT-4o?
A power user would need to dive deeper to notice the differences between GPT-4.5 and GPT-4o, but a closer look reveals some impressive improvements. In tests, GPT-4.5 outperformed GPT-4o in poetic challenges, providing more evocative descriptions of rain-soaked New York City scenes.
The model also excelled in subjective follow-up questions, offering explanations that felt more human-like and empathetic. For example, when asked about Michelle Yeoh’s Best Actress Oscar win, GPT-4.5 provided a thoughtful analysis that covered the actress’s performance and cultural significance.
GPT-4.5’s ease of use was also demonstrated in simple prompts like cooking recipes. In contrast to its predecessor, the new model provided a straightforward and easy-to-follow solution for a honey garlic salmon recipe.
However, subtle differences may be lost on casual users. GPT-4.5’s emotional intelligence is still not as pronounced as some might expect, and its poetic output can be somewhat predictable. Nevertheless, for those willing to explore beyond the surface level, the improvements offered by GPT-4.5 are undeniable.
For now, users on the free tier of ChatGPT Plus won’t miss out on much from not having access to the new model. As GPT-4.5 is still a research preview and not fully complete, further experimentation may reveal its full potential.
Source: https://www.techradar.com/computing/artificial-intelligence/chatgpt-4-5-understands-subtext-but-it-doesnt-feel-like-an-enormous-leap-from-chatgpt-4o