OpenAI have announced the o3 ChatGPT model is launching in January. They can’t call it o2 because of o2 the phone company…
O3 is a huge huge leap on the ARC-AGI test compared to the previous best model o1.
This test is widely accepted as being the standard benchmark test for if artificial general intelligence has been achieved or not.
Arc agi details on the link below.
as you can see from the graph, models were starting to flatline and experience diminishing returns, so this is a huge moment in history. As for the arc prize? They have announced a new V2 test that o3 can only score 23% on. This is kind of moving the goalposts though.
Created at . Page last updated at .