XARIV
← Workflow

Step 2 · Benchmark

XARIV Pulse

Inference Benchmarking & Telemetry

See the latency profile before you serve a single request.

Point Pulse at a public dataset like ShareGPT or paste your own prompts. It replays the request mix on your target GPU and summarizes TTFT, ITL, TPOT, end-to-end latency, throughput, and live-style GPU telemetry.

Product overview →← Define workload (Lens)Export decision report →

Profile a workload

Pick a dataset and target hardware — Pulse replays the request mix and reports the latency profile and GPU telemetry.

Real multi-turn ChatGPT conversations — long, varied prompts and answers.

Choose a dataset and hardware, then run a profile to see the inference metrics.