Models
Mixture-of-Experts vs Dense Models: Which Fits Your Hardware
MoE gives you a small model's speed with an enormous model's memory footprint. That tradeoff decides more deployments than quality does.
#inference
2 articles2 articles on Model Drop tagged "inference."
2 articles
-
-
PlatformsServerless GPU Platforms for Inference: Reviewed
How pay-per-second GPU platforms actually perform on cold starts, scaling, and cost compared to reserved capacity.