Efficiency Index

GPU cost and speed on replayed agent workloads. Measured and simulated points are marked.

LLM serving benchmarks.

Throughput, latency and cost per token by GPU, on chat and agent workloads.

Models

Workloads