OpenAI announced the new language model GPT-4.5. Although the company says that this model is the “largest and most knowledgeable” model they have developed so far, it will not bring a revolutionary innovation and may not be as strong as O1 or O3-Mini.

The GPT-4.5 will have more advanced spelling skills and expanded data set, but it is not a model that can be called “pioneer” compared to OpenAI. It was stated that GPT-4.5 increased the calculation efficiency of 10 times in the leaked internal documents, but it would not have new features at the expected level.

Error Ratio Lower than GPT-4O

In the training of GPT-4.5, OpenAI’s O1 is reported to be used by Strawberry.

The company says that this new model has trained this new model with techniques, such as SFT (SFT) and human feedback -supported learning (RLHF). The error is obviously low than the GPT-4O, but it will be slightly higher than the O1 model.

GPT-4.5 will understand your feelings more enough

GPT -.5 will be able to understand your intentions more appropriate from your voice tone when you are under the meaning of the users in the conversations.

For example, if you tell the GPT-4O after a test you have passed through a period of power, you will list the things you can do to beautify your mood. GPT-4.5 will understand that you don’t want anything, even if it’s just a chat boat, you’ll want to hear a little understanding and soothing sentences.

Chatgpt Pro subscribers, OpenAI’s 200-dollar monthly plan, will gain access to GPT-4.5 at Chatgpt as a segment of research preview. Developers in the priced subscription options of OpenAI’s API will also be able to use the GPT-4.5 from today.

Repetition of the publication in which GPT-4.5 is introduced:

The leaks that emerged at the beginning of this week claimed that the GPT-4.5 would be presented to the operator model just to the Chatgpt Pro users. OpenAI will soon make an announcement to share details of when the model will offer all users to use.