# ray-project/llmperf

LLMPerf is a library for validating and benchmarking LLMs

Repository: https://github.com/ray-project/llmperf
Canonical: https://ross.abutalabs.com/products/llmperf
Language: Python
License: Apache-2.0
License Family: permissive
Archived: true
Last push: 2024-12-09T01:52:28+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 75
- inputs: {"age_days": 1058, "days_push": 633, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1127, forks 203 (observed 2026-08-28T04:03:41.656090+00:00)

## What it is
LLMPerf is a Python library for benchmarking and validating the performance of LLM APIs. It runs load tests measuring inter-token latency and throughput, plus correctness tests, across providers like OpenAI-compatible endpoints, Anthropic, TogetherAI, and Hugging Face.

## Use cases
- benchmark llm api latency and throughput
- load test an openai-compatible endpoint with concurrent requests
- compare performance of different llm providers
- measure inter-token latency for streaming llm responses
- validate llm api output correctness
- stress test llama models on hosted endpoints

## When to choose
- you need standardized performance metrics (TTFT, inter-token latency, throughput) for LLM APIs
- you want to compare multiple LLM providers with consistent token counting
- you are evaluating inference endpoints before production deployment

## When to avoid
- you need to benchmark self-hosted model hardware rather than APIs
- you need functional/quality evaluation of model outputs beyond basic correctness checks
- you need a actively developed tool with recent features

## Facets
- artifact type: library
- maturity: maintenance
- function: benchmarking, llm-inference, testing
- domain: large-language-models, performance, machine-learning
- platform: python, cli
- tags: llm-apis, load-testing, latency-measurement, throughput, openai-compatible

## Member repositories
- ray-project/llmperf (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:41.656090+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:38:53.355619+00:00, confidence not recorded.
  - readme: https://github.com/ray-project/llmperf (fetched 2026-08-28T04:03:41.656090+00:00, sha 066e6f0e195e)
- Data as of 2026-08-30T08:39:29.467469+00:00.