San Francisco: Recently, OpenAI has stepped up the artificial intelligence race by deploying two groundbreaking models of O3 and O3 Mini which all point towards a possible development of an AGI.
Taking part in the company’s 12-day ‘Shipmass’ event, Sam Altman is focused on presenting these AI models as notable improvements over their O1 counterparts. That launched is a direct competition to search titan Google’s capabilities especially with the unveiling of ChatGPT Search at the same time.
Key Highlights of the New Models:
Performance Breakthroughs:
O3 is claimed to have attained a 22.8% success rate of coding tests with SWE-bench verification.
Solved 25.02% of all the high level of math and reasoning problems
Scored higher than most of its precursors and Gao, OpenAI’s chief scientist in code competitions
Mathematical Prowess:
Won the AIME 2024 contest by solving 87.7% of the scientific problems of an expert level
Generally achieved much higher accuracy than that of other AI models
Cautious Method when it comes to Utilisation
Citing potential risks, Altman has been calling for a set of federal guidelines before broad testing. These models are currently being solicited for applications from Researchers who, by 10th January 2025, will competitively validate these models.
Potential Launch Timeline:
O3 Mini: Launched before the end of January 2025.
Full O3 Model: Expected to be launched later the year after Mini version