Chubby's GPT-5 Review (via)
“All in all, I would say that GPT-5 is exactly what we expected. No more and no less. Outstanding evaluations, probably the best AI model in the world, and optimized exactly where it was needed. I look forward to testing it myself. If necessary, I will update my review. Great work, OpenAI!”
All I can say is: really? Really?
After all that time what you were expecting was a minor performance bump compared to other current SOTA LLMs? Does anyone remember when GPT-4 was released? I remember when GPT-4 was released nothing even came close for over a year. Even when other AI labs figured out how to build the original GPT-4, OpenAI had already released a half dozen new upgrades and improvements that kept them at the top of the benchmark leader boards for what felt like an eternity. Everyone went crazy when Anthropic’s Opus beat GPT-4’s dominance for a few days before OpenAI released another update.
It feels like for some reason all the AI influencers have to pretend this was a revolution because otherwise they’d all feel a bit silly for hyping this model release so hard.
GPT-5 was not “exactly what we expected”. No one can look at this graph of people’s expectations on betting markets for LLMs and say with a straight face that it’s exactly what we expected.
