I'm really perplexed with the latest OpenAI GPT-4.1 release.
I just don’t understand the goal behind releasing an obviously inferior model.
The naming is pretty terrible—half of the hacker news comments is about that.
But the bigger issue is that the new model is clearly worse than many other non-OpenAI models.
SWE bench puts it on the third spot. The aider leaderboards don't even rank it in the top 10 with 52.4% correctness. The first spot (Gemini 2.5 Pro) is both smarter and cheaper (72.9%, $6.32)!
And in this light, the latest news that OpenAI is working on its own social network doesn't exactly inspire confidence. As if, we can't build a competitive model anymore, so we are going to build a Twitter, instead.