OpenAI officially launched the large-scale natural language model GPT-4.5, claiming that it can communicate with users in a more natural and warm form.
It can reduce the situation of answering irrelevant questions and talking nonsense seriously

Previously announced that it would launchThe last non-chain thinking model GPT-4.5Afterwards, OpenAI officially launched this large-scale natural language model earlier, claiming that it can communicate with users in a more natural and warm manner.

OpenAI said that GPT-4.5 will be able to give more appropriate responses based on the content of user interactions, rather than providing unnecessary suggestions. It emphasized that this is because GPT-4.5 has more complete knowledge and understanding of the world, so it can reduce the situation of answering irrelevant questions and talking nonsense in a serious manner.

Furthermore, GPT-4.5 achieved a 37.1% accuracy rate on OpenAI's SimpleQA test, surpassing GPT-4o, o3-mini, and o1. However, GPT-4.5 currently lacks multimodal support and is not suitable for applications such as voice interaction, video analysis, or image sharing. Furthermore, it is intended for broader intelligent model applications rather than replacing existing models.

GPT-4.5 is currently priced at $75 per million words and is currently only available to paid ChatGPT users. It will also be available to users of ChatGPT Plus and ChatGPT Team plans in the future.