Directionally that is the goal, funding local inference
For now we are testing TEE with a few dozen people on a beast setup that we will soon upgrade to Kimi 2.7
Meanwhile you can do a surprising a...
I’ve got local GPUs with almost as much memory as a Spark (but way more BW). I was mostly tongue-in-cheek referring to the annoying reduction in 120B-class models lately and the increasing reality that not many can afford to access frontier models lately.
2