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.
#self-hosting
4 articles4 articles on Model Drop tagged "self-hosting."
4 articles
-
-
ModelsOpen-Weight vs Closed Models: The Decision Isn't Capability
Benchmark parity is real for routine work. The choice turns on residency, version pinning, and where your cost curves cross.
-
ModelsSmall Models That Run on One GPU, and What They Cost You
A quantized 27B model fits in 24 GB and handles most routine work. Throughput, not capability, is what actually limits it.
-
PlatformsSelf-Hosting vs API Inference: Utilization Decides It
Free weights are not free inference. The cost case lives on a utilization curve, and residency is a better reason anyway.