@ravisparikh Don’t take it personally. I don’t have any reason to think yours in particular are fake. But I have seen too many fake ones to believe any of them.
@theo Ok but how much does a subscription get you? Because I am too poor for anything above $20 a month, and I am pretty happy with the usage Anthropic gives for Opus 5.5 Medium. Is Neon on $20 comparable?
@JohnMcKeownn@thsottiaux Nah, humans don’t ask questions that only they would understand, using names they invented in their own thinking. Humans have theory of mind, which prevents that kind of thing.
@TokenGremlin You doubt Opus 5.5 has been surpassed, yet put Google at number 1. Bro I think I need to just mute you, your takes are always hot ass garbage
@thsottiaux “We understand that our language model can’t speak language, but it can design a GPU for you! And we know how many of you out there build GPUs at home, so we are prioritizing that over all of those weirdos who use the language model for language.”
@EugeneSoloviov@blingdivinity Yeah their numbering/naming system makes no sense. They are obviously using a different scheme than other companies, so the suggesting that a different name is a different size/tier just doesn’t apply to them.
@55kkeett@blingdivinity Because people build workflows around specific models that don’t necessarily work as well with a new model, even if the new model is better. It will get depreciated eventually, with advanced warning.
@thsottiaux@MiaAI_lab I think we might have skipped the part you are supposed to do after getting feedback, because these models (at least 6.2 Sol, haven’t tried Astra) are bad.
@thsottiaux Doesn’t at all explain the Fable/Astra comparisons people are posting from today. It is the difference between a kindergartener and a postgrad.
@thsottiaux So far, my impression of 6.2 Sol has been meh. Still consumes way too much usage for simple tasks when compared to Opus 5.5 and has to be handheld every step of the way or else it makes awful design decisions.