yes, let me ignore all the scientific measurements of capability the world has to offer and measurements of kl-divergence and analytics, and trust John Pener who was in China.
@sabaka_339@VedikaAPI@OpenAI I don't know whether you have ever used a Chinese model. I know they are cheaper, but they are not as good as those benchmarks make out them to be. I was in china, I used the Xiao Mi models. Trust me, you cannot build with them.
@TheAhmadOsman i respect your mission but these kind of posts are what remind me that this social network and maybe social media in general are a hopeless source of noise
I’m seeing so many of these posts, and my experience is the same
If Sol 6.1 runs at sol 5.6 speed, the usage drains roughly the same as before or worse
They didn’t fix shit, it’s still broken usage just disguised
@thsottiaux@hiarun02 usage is an abstract quantity. we need a verifiable unit of usage so we can make sure we are getting the quantity/dollar we expect. if that unit changes, fine, but we have to have full transparency. this isn't it.
@thsottiaux you're not "locking in" while 6.1 sol is still unusably slow. @OpenAI how about you lock in an apology and some resets to compensate for this nonsense?