Hi nvidia team. Glm 5.2 weights have been released can we have the model on the api ?
Never say neeeveeeer.
why are you saying that?
What?
i hope that nvidia will remove GLM5.1 to add the awsome GLM 5.2 with 1M context <3 GLM-5.2 (max) - Intelligence, Performance & Price Analysis
fah, no need to remove glm 5.1
of corse they need, do you think that the hardware is unlimited ?
more model doesnt consume more hardware, hardware usage is determinated by user usage. If no one use GLM-5.1 then no hardware will be sonsumed even if it exist.
are nvidia staff ? you know how itās work ? so can you explain why minimax or ds4 pro/flash are really slow and why glm 5.1 is really fast ? maybe because they not shared the same hardware no ? so no all ia are not using the same hardware, so they have to release ressource to add use them on other model seem logical
so maturity in you language ššš
from what i can see, they quantized glm 5.1. Iāve been getting some spelling mistakes and the reasoning process is a lot shorter than it was around a week or two ago, before the models started throwing 429 errors. thus why itās faster now, it was way slower some time ago.
this guy bozoweed likes arguments
Nvidia basically has unlimited hardware resources right now. They quantize models specifically for their own hardware. This allows them to deliver a turnkey āmodel + hardwareā bundle, which customers can confidently purchase after testing it out on Nvidiaās infrastructure
intersting i donāt even know that they quantized glm5.1, but that explain what is going faster now , thx