Scaling past a trillion parameters requires inference engineers to solve 3 issues:
- Width: Routing between hundreds of sparse experts
- Depth: Mixing global and local linear attention
- Length: Processing 1M-token sequences
I asked Ali and Bryce to draw it up. They delivered.
@blader We found that you are an idiot in our internal benchmarks one month ago, but now we find you twice an idiot in our internal benchmark pro.
Honestly we were not expecting that, congrats.
@notsurajgaud Why are you charging so less, for the absolute AI slop.
This is a scam, you are selling courses from AI to inference, without even shipping a single AI future to production.
Please charge 100k dollars per month.