Hey guys, I was hoping for a bit more news this week, but I guess this stuff is still exciting and I've waited long enough.

OpenAI:

o3 is here! ...Kinda. If you remember a few months ago we discussed o3, OpenAI's next gen high performance chain-of-thought (reasoning) model. This is a model that was so good at reasoning that it technically qualified as Artificial General Intelligence (AGI). However, the standards have been recently redefined as the real world performance of this apparent "AGI" is much lower than would be expected.

All that however was for the big version of o3. The one that just came out is in fact the mini version, which is much weaker. It does still beat o1 on most benchmarks, putting it top of the class and in close competition with DeepSeek R1.