---
title: "What 38 RAG systems reveal about enterprise retrieval"
description: "38 systems on one enterprise RAG benchmark: a 3.7x score spread, default configs far behind, and which leaderboard gaps are actually noise."
canonical: https://annamlabs.com/blog/what-38-rag-systems-enterpriserag-bench-reveal
---


# What 38 RAG systems reveal about enterprise retrieval

38 systems on one enterprise RAG benchmark: a 3.7x score spread, default configs far behind, and which leaderboard gaps are actually noise.

*Võ Ngọc Quang Minh · October 31, 2026 · 12 min read*









| codos.ai | 92.64 | 94.4 | 95.16 | 96.45 | 0.62 |
| --- | --- | --- | --- | --- | --- |
| Flexor.ai (+ GPT-6.1 Sol) | 91.48 | — | — | — | — |
| CDL (Causal Dynamics Lab) | 90.91 | — | — | — | — |
| AHOY Labs | 88.26 | — | — | — | — |
| Skyller AI | 87.16 | — | — | — | — |
| OpenAI File Search | 61.03 | — | — | — | 15.7 |
| BM25 + GPT-5.4 (one-shot) | 50.6 | — | — | — | — |
| RAGFlow | 50.24 | — | — | — | — |
| Amazon Q (Kendra) | 48.96 | — | — | — | — |
| Azure AI Search | 48.42 | — | — | — | — |
| Vertex AI Search | 41.87 | — | — | — | — |
| LlamaIndex (default config) | 27.2 | — | — | — | — |
| LangChain (default config) | 24.98 | — | — | — | — |
| Open WebUI + Chroma | 24.89 | — | — | — | — |
















---

Read the HTML version with visuals: https://annamlabs.com/blog/what-38-rag-systems-enterpriserag-bench-reveal
More articles: https://annamlabs.com/insights