# C15 — Introducing the MLPerf End-to-End RAG Inference Benchmark

- Publisher: MLCommons
- Source: [Introducing the MLPerf End-to-End RAG Inference Benchmark](https://mlcommons.org/2026/08/endtoend-inference/)
- Published: 2026-08-26
- Updated: Not stated / not established
- Accessed: 2026-10-22
- Type: Official benchmark announcement
- Evidence: opened primary page; relevant content read.

## Paraphrased finding

RAG benchmark measures ingestion in documents/second and QnA in tasks/second; its first version is Offline, with all requests available together.

## Limits

Offline throughput is not online user experience or tail-latency evidence.