@thsottiaux Once I saw a similar message like this on X and they came back saying it was a “skill” issue 🥲… please don’t let us down in the same way @thsottiaux
@AnthropicAI In other parts of the globe if a leader was to “*dictate*” what a private company could or couldn’t do, we would have a specific word to define such “leader”
@mattshumer_ In other parts of the globe if a leader was to “*dictate*” what a private company could or couldn’t do, we would have a specific word to define such “leader”
Running it on my RTX5090, the thing is great. It does about 80% of what Opus is capable of (as a ballpark estimate, of course)… Nevertheless, if we take into consideration how much Anthropic nerfed Opus and rating limits, being 20% inferior sort of makes it nearly like-for-like
Running it on Q6 with KV cache Q4_0, context size of 262k, with a parallelism of 2
@TheAmolAvasare Doubling limits? Please stop it, you know this will actually just throw people faster into a full extra usage territory sooner (and those who doesn’t reach it will just be managing to do so now)
@YazSec@TheAmolAvasare They are not… in reality this sounds like a strategy to push people faster into a full extra usage territory… amazing how they tried to wrap it up as a great thing
@rezoundous If instead they could fix whatever they did to mess up so much token consumption I bet it would be way more impactful in any reputation and market share getting lost