Skip to main content

Comparison

This is a scoped comparison of client-side inference measurement tools, compiled from public project pages and documentation as of September 2026. It is not a hands-on bake-off. Metric names only match when tokenizer, prompt sequence, sampling parameters, warmup, measurement window, endpoint topology, and SLO definitions match.


What Bench ships at 1.0

CapabilityWhere
Four OpenAI-compatible modalities: LLM, VLM, ASR, imagegenmetrum-ai-bench {llm,vlm,asr,imagegen}
Open-loop scheduling with seeded Poisson (or constant) arrivals--request-rate, --arrival, --seed
Pooled multi-endpoint distribution--endpoints-file; aggregates labeled pooled_mixture
TTFT = first visible output token (not first-byte headers)ttft_s; headers are first_byte_s
Warmup excluded from measured distributions--warmup-requests
Goodput requires stated SLOs--slo ttft=|tpot=|e2e=; without SLOs, goodput equals validity-filtered throughput on strategic
Partial-run flagCtrl-C / SIGTERM drain writes partial: true
p99 flagged unreliable at low np99_unreliable when n < 100
Hyndman–Fan type 7 percentilespercentile_method: hyndman_fan_type7
Concurrency/rate sweeps, HTML report, MLPerf-shaped exportstrategic: --sweep, --html, --mlperf-dir

Landscape

ToolWhat it answersOverlap with BenchWhat Bench addsWhat it does better than Bench
SemiAnalysis InferenceXHow named stacks score on a published, recurring hardware CI matrixClient-visible latency/throughput languageLocal, operator-owned runs with a stamped manifest; four modalities in one clientContinuous public CI and a recognized public scoreboard
NVIDIA AIPerf (successor to GenAI-Perf)How a Triton / NIM / NVIDIA-oriented path behaves under loadRate, Poisson-style arrivals, tokenizer counts, SLO-style ideasMulti-modality and pooled multi-endpoint mixture labeling; strategic sweeps/exportsEcosystem familiarity and telemetry already wired into many NVIDIA pipelines
MLPerf InferenceWhether a submission meets audited workload and accuracy rulesLatency/throughput language; Server/Offline namesInformal LoadGen-shaped export onlyAudited rules, accuracy requirements, submission standing
vLLM / SGLang bench_servingHow this engine build behaves on these datasetsRate/concurrency, seeds, ignore-EOS, local tokenizer countsCross-engine OpenAI-compatible client; multi-modality; strategic exportsAlready installed next to the engine; deep stack-specific datasets
GuideLLM / LLMPerfLatency/throughput under scripted loadRate/concurrency sweeps; HTML-style reporting (GuideLLM)Shared Rust client across four modalities; typed errors; stamped summary.v3Packaging / Python ecosystem fit for quick scripts
lm-evaluation-harness / HELM / OpenCompassQuality and task scoresOccasional shared prompt corporaBench measures serving performance (plus ASR WER/CER only)Breadth of quality tasks and academic protocols
NVIDIA DynamoHow to run multi-node inferenceYou may point Bench at Dynamo-fronted endpointsDynamo is a target, not a measurement alternativeEnd-to-end serving product features

MLPerf export disclaimer

metrum-ai-bench-strategic --mlperf-dir … writes LoadGen-shaped text for parser-oriented interoperability. It is not an audited submission. Exports begin with an UNOFFICIAL disclaimer and refuse the official LoadGen substring Result is : VALID.


What we do not claim

  • Not first, only, or industry-standard.
  • A stamped manifest is table stakes for publishable runs, not a moat.
  • Not a competitor to MLPerf Inference submissions.
  • Not a head-to-head performance bake-off against AIPerf or InferenceX.
  • No claim that Bench replaces engine-native bench_serving for engine developers who already live in that tree.

Not in 1.0 (roadmap)

Planned for a next release (no dates):

  • Agent mode
  • Quality metrics beyond ASR WER/CER
  • Cost per accepted task

Bench vs Bench Platform

Metrum AI Bench CLI (this product)Metrum AI Bench Platform
RoleMeasure one environment; emit a result with a manifestRemember, compare, govern, and attest
LicenseApache-2.0 clientCommercial
DocsThese pages cover the OSS CLICommercial product; not covered here - contact Metrum AI