is it actually usable for models that require that much RAM or is it still bottlenecked by the RAM size of the largest individual machine in the pool?
as in, can one actually run a model that needs this much or is it just a big number that doesn't translate to practical benefits directly
Models are still limited by the per-machine RAM sadly, but we're *thrilled* to see that there's this much activity already, which will translate to redundant nodes and parallelization of workloads