The American company OpenAI has announced the release of a new version of its artificial intelligence model-GPT-4.5.
"We are releasing a preliminary version of GPT-4.5 — our most powerful and advanced model. It significantly improves the ability to understand hints, make logical connections, and generate creative ideas, " the company said in an official statement.
The model is optimized for more accurate processing of user queries, which makes it particularly useful in areas such as text writing and programming. OpenAI developers expect that GPT-4.5 will allow fewer errors compared to previous versions.
OpenAI emphasized that GPT-4.5 is an example of scaling unsupervised learning. The model was trained on Microsoft Azure supercomputers and has a deeper understanding of information. It is expected that this will help reduce the level of so-called "hallucinations" of AI and increase the reliability of responses.
Earlier, The Verge reported that the new model was developed under the code name Orion and will be the latest version of OpenAI, based on the traditional architecture. In the future, the company plans to introduce GPT-5, which is scheduled for release at the end of May.
At the moment, the most advanced model remains the GPT-4, introduced in March 2023. It is capable of processing both text and visual information, demonstrating a high level of accuracy in professional and academic tests. According to OpenAI, ChatGPT-4 is 82% less likely to respond to requests for prohibited content and 40% more likely to give correct answers compared to ChatGPT-3.5.
Despite significant improvements, the model is still prone to errors and may contain inaccuracies in some statements.
