MEASURED MODEL SERVING RESULTS

Inference workloads

Explore measured deployments by GPU family and cluster size. Benchmark matches describe the hardware used in a run; they do not verify a provider offer’s memory or interconnect fit.

SemiAnalysis InferenceX · AgentX

CURRENT AVAILABLE OFFERS

Choose a GPU cluster

Loading offers…

GPU familyModelMinimum P90 interactivity

tok/s/user

Best blended rates use a 5:1 output-to-input price ratio, full measured throughput, and GPU rent only.

Loading available GPU clusters…