ChatGPT is getting even smarter. OpenAI has unveiled its latest model, GPT-4.5.
Research lead Mia Glaese revealed that, starting today, the new model will be available as a research preview to ChatGPT Pro users and developers in the API.
The model will make its way to ChatGPT Plus users from next week, according to Glases.
OpenAI says GPT-4.5 is its most knowledgeable model yet, bringing advancements in unsupervised learning and reasoning. The improvements should be especially helpful in handling science-related queries. You can also expect boosts in word knowledge and intuition, along with reduced instances of hallucination — occasional bugs that offer completely made-up, inaccurate information.
Compared to previous iterations, GPT-4.5 also comes with better contextual knowledge which will supposedly make it more useful for writing, programming or practical problem-solving.
In a demo comparison against GPT o1, the new model was capable of picking up on social cues in user questions involving personal situations, like responding to a friend who recently cancelled plans.
In the demo, in addition to an emotionally charged, somewhat angry response, GPT-4.5 also offered a more tempered response compared to o4. That’s thanks to its new design, which aims to make it more human and fine-tuned to emotional nuances.
Indeed, OpenAI says testers ranked it higher in that regard compared to previous models.
How Will GPT-4.5 Change ChatGPT
As already mentioned, the biggest changes we can expect in ChatGPT thanks to the new model is a boost in its emotional intelligence and a more natural, conversational approach to answering questions.
It also appears the new model outperforms previous models in multilingual processing, which should make it more useful to non-English speakers.
GPT-4.5 is already rolling out to ChatGPT Pro users on web, mobile and desktop, with Plus users expected to get the upgrade by next week.
