The latest Kimi model, which scores higher than 5.1 Fable on Artificial Analysis, is estimated to require around 6-7 TB of VRAM to run. If we are going with many A100 GPUs for this, then we are looking around a million dollars in GPUs alone
149 karma · joined October 7, 2025
The latest Kimi model, which scores higher than 5.1 Fable on Artificial Analysis, is estimated to require around 6-7 TB of VRAM to run. If we are going with many A100 GPUs for this, then we are looking around a million dollars in GPUs alone
This is not exclusive to the US, though. NordVPN is an example - it launched in Lithuania and now has global reach, and thousands of EU companies have went that path as well. I am not saying that it is on par with the US on this metric, but let's not pretend that such things don't happen in the EU
Subtle
By this they probably mean RTX series GPUs? If so, then they are not comparing the hardware efficiency with the A100 / H100, etc. that are commonly used for training models
It can be very hard to prove in court that your data was used for training
Perhaps, but not everyone carries their phone around with them all the time
Uber drivers are generally a joke compared to black cab drivers in pretty much every way
Curious, do you have a source for this apart from Apple's marketing?
Metal Gear Solid did this 1998
Great video about this article from a political science / game theory professor: https://www.youtube.com/watch?v=U-ewrIenEbo
Wars are not contests of hypothetical maximum destructive capacity. They are attempts to obtain political goals at an acceptable cost. It was a strategic and political defeat, even though it was not an unconditional military surrender.