OpenAI's 12-Day Series has finally ended, and on the last day, they announced the O3 and O3 Mini models, which have smashed all benchmarks! 1. O3 scored 2727 Coding Elo on Codeforces, ranking it equivalent to the 175th best coder globally. 2. On Hard Math problems for which even experts take hours or days to solve, O3 achieved a remarkable 25% result, while other AI models barely reach 2%. 3. Most intrestingly, On the Agi benchmark by ArcAGI, O3 scored an incredible 87.5%, far surpassing GPT-4's 11%. Also Sam Altman confirmed these models will launch in late January. It's really surprising how fast OpenAi is shipping these products and how crazy they are getting! Btw What do you think, how far are from agi?
Download the medial app to read full posts, comements and news.