Openai GPT-4.5 announces, warns that it is not a model of AI border

Openai is GPT-4.5 launch todayThe latest and largest form of artificial intelligence language. GPT-4.5 will be available as a research inspection for Chatgpt Pro users to start. Openai calls the version “its most familiar with it”, however Initially This GPT-4.5 is not a model of borders and may not lead as well as O1 or O3-MINI.

GPT-4.5 will get better writing capabilities, improve global knowledge, and what Openai “refined personality” will have previous models. Openai says that interacting with GPT 4.5 will feel “natural”, adding that the model is better in identifying patterns and drawing contacts, making it ideal for writing, programming and “solving practical problems.”

However, Openai notes that it will not provide sufficient new potential to be considered a model of the border. “GPT-4.5 is not a model of the border, but it is the largest LLM in Openai, and it improves the arithmetic efficiency of GPT-4 with more than 10x,” Obray said. In a document That leaked before it was announced. “7 new border capabilities do not offer compared to previous thinking versions, and their performance is less than O1 and O3-MINI research and deep research on most of the readiness assessments.” Openai has since removed this signal from An updated version From the document.

It was previously reported that Openai was using its O1 thinking model, Codenamed Strawberry, to GPT-4.5 training With artificial data. Openai says he trained GPT-4.5 “using new supervision technologies along with traditional methods such as SFT and learning reinforcement from human comments (RLHF), similar to those used in GPT-4O.”

Despite some of its own restrictions, the GPT-4.5 is much less than the GPT-4O, according to Openai, and a little less O1. “We have aligned GPT-4.5 to be a better collaborator, making the conversations feel warm, easier and emotionally delicate,” Rafael Guntego Lopees, Openai, said during Livestream. “To measure this, we asked human beings to be evaluated against GPT-4O, and the GPT-4.5 superior to each category mainly.”

In a post on xSamani, CEO of Openai, acknowledged that GPT-4.5 is a “giant and costly model” and that “it will not crush the standards.”

After launching the Pro us users, Openai says that GPT-4.5 will be offered to Plus and team users next week and then to Enterprise and EDU users after that. It’s too Available in Microsoft’s Azure Ai Foundry Plastic, along with new models of stability, Cohere, and Microsoft.

We revealed in Noteman Last week This Openai was planning to launch the GPT-4.5 by the end of February and GPT-5 in late May. CEO of Openai Sam Altman GPT-5 is referred to As “a system that integrates a lot of our technology”, and will include Openai’s New O3 Thinking FormThe company disturbed the 12 days of Christmas ads in December.

While openai O3-Mini Sadr Last month, Openai only offers O3 as part of the upcoming GPT-5 system. This is in line with Openai’s goal in combining its large language models to create a more capable model that can be described as artificial general intelligence, or AGI.

Update, February 27: Note that Openai changed the GPT-4.5 system and added information from Sam Altman.