On the last day of the twelve-day product release live broadcast, openai released a big move, the finale "OpenAI o3". The ability of o3 is almost a direct dimensionality reduction attack on all current models.
A few reviews I saw:
1. Figure 1 Software Engineering Test (SWE-Bench Verified), this is like a test for writing programs, for example, if you write a software, it must be fast, accurate, and not have bugs (small errors). This is an examination of o3
twitter.com/qinbafrank/status/...