|
|
|
|
|
|
by traes
23 hours ago
|
|
|
Presumably it's a rounding error compared to their full output, and they're making sure they have enough compute set aside for research by limiting public models. The more datacenters they build the less they have to limit them. |
|
|