Igeekphone February 28 news, this morning OpenAI held a live event, officially released the latest large model GPT-4.5, will be open to paying users in stages.
ChatGPT Pro users will be able to use GPT-4.5 on the web, mobile, and desktop versions starting today, and will be available to Plus and Team users next week, and to enterprise and Edu users the following week.
GPT-4.5 is the largest chat model OpenAI has ever built, and it is more than 10 times more efficient than GPT-4.
Altman tweeted that GPT-4.5 made him feel like he was talking to a thinking person for the first time and could get really good advice from the model.
GPT-4.5 is a very large and computation-intensive model with an API price of $75 / million tokens input and $150 / million tokens output, which is 30 times the price of GPT-4o, and OpenAI says it is evaluating whether to provide it in the API in the long term.
Of course, the large scale also brings superior performance, GPT-4.5 in the SimpleQA benchmark accuracy of 62.5%, hallucination rate of 37.1%, better than the previous model.
Better ability to understand human intentions, higher “emotional intelligence”, more natural response, can play a greater role in improving writing, programming and solving practical problems, and the model has fewer hallucinations, but the mathematical reasoning ability is still inferior to inference models such as o3-mini.
GPT-4.5 is OpenAI’s last “non-chain of thought model,” which improves unsupervised learning by scaling up pre-training and post-training, and will serve as the basis for inference models in the future.
However, GPT-4.5 only supports search, upload files and pictures and canvas functions, and does not support multimodal functions such as voice mode, video and screen sharing, and OpenAI says it will continue to update in the future.