# Atlan vs Amundsen: which do AI models recommend for data catalogs, October 2026

IT AI Recommendation Index, October 2026 Edition, Data catalogs and governance. Ten of fourteen models named Atlan first on the direct prompt; zero named Amundsen. Page: https://it-ai-index.com/it-data/data-catalogs-and-governance/atlan-vs-amundsen/

| | First-choice share | Rank | Negative rate | Labels | Models naming it |
|---|---|---|---|---|---|
| Atlan | 44% | #1 of 15 | 10% | 58 | 14 of 14 |
| Amundsen | 4% | #7 of 15 | 33% | 12 | 7 of 14 |

## The direct prompt, model by model

- Claude Haiku 4.5: atlan first (first choices: Atlan) (alternatives: Alation, OvalEdge)
- Perplexity Sonar: atlan first (first choices: Atlan) (alternatives: Alation, OvalEdge, Select Star)
- Grok 4.1 Fast: atlan first (first choices: Atlan) (alternatives: Alation, OvalEdge, Secoda)
- Mistral Small: atlan first (first choices: Atlan, OvalEdge)
- Llama 4 Maverick: atlan first (first choices: Atlan)
- Qwen 3.7 Flash: atlan first (first choices: Atlan) (alternatives: AWS DataZone, DataHub, Google Dataplex, OvalEdge, Seceda)
- GLM 4.7 FlashX: atlan first (first choices: Atlan, Secoda, Select Star) (alternatives: Informatica Intelligent Data Management Cloud, OvalEdge, data.world)
- MiniMax M2.5: atlan first (first choices: Atlan, Secoda) (alternatives: Alation, Microsoft Purview)
- GPT-6 Luna: atlan first (first choices: Atlan) (alternatives: Microsoft Purview, OpenMetadata, Secoda)
- Muse Glimmer 30B: atlan first (first choices: Atlan) (alternatives: Alation, OvalEdge)
- Gemini 3.5 Flash: neither first, one named (first choices: Select Star) (alternatives: Atlan, Coalesce Catalog, OvalEdge, Secoda)
- DeepSeek V4 Flash: neither first, one named (first choices: Secoda) (alternatives: Atlan, DataHub)
- Kimi K2: neither first, one named (first choices: Secoda) (alternatives: Atlan, OvalEdge, Select Star)
- GPT-5.4 mini: neither named (first choices: DataHub) (alternatives: Alation, DataGalaxy)

## What the models said about Atlan

- "Options to Avoid ... Commercial tools (Collibra, Alation, Atlan, Select Star)" (DeepSeek V4 Flash, budget prompt, hard negative)
- "Avoid pricier ones like Atlan/Alation ($10k+/year)." (Grok 4.1 Fast, budget prompt, hard negative)
- "a lower-tier hosted catalog like Atlan or DataGalaxy can reduce ops burden, but these are usually not the best fit for a tight budget" (GPT-5.4 mini, budget prompt, soft negative)
- "Atlan is cited as: Gartner recognizes Atlan as a Leader in both the Metadata Management Solutions Magic Quadrant (2025) and, as of 2026, the Magic Quadrant for Data and Analytics Governance Platforms" (Muse Glimmer 30B, comparative prompt, first choice)
- "Atlan is the leading active metadata platform in the current generation of the category and the most commonly shortlisted catalog for cloud-native data stacks in 2026." (Claude Haiku 4.5, direct prompt, first choice)
- "Some modern entrants (like Atlan) bridge this gap by offering strong automation but allowing developers to extend functionality via API." (Qwen 3.7 Flash, scale prompt, first choice)

## What the models said about Amundsen

- "Avoid for Mid-Market: ... open-source like DataHub/Amundsen (free but needs engineering overhead)" (Grok 4.1 Fast, direct prompt, hard negative)
- "Amundsen (Lyft) — archived and no longer maintained as of 2026, so don't build on it" (DeepSeek V4 Flash, budget prompt, hard negative)
- "have lower license costs but require significant DevOps effort to install, maintain, and keep running" (Qwen 3.7 Flash, scale prompt, soft negative)
- "likely to be an open-source tool such as Amundsen or DataHub, which are suitable for on-premise environments" (Llama 4 Maverick, budget prompt, first choice)
- "Best ultra-lean choice: Amundsen" (GPT-5.4 mini, budget prompt, first choice)

Share is the count of first choices across the direct, paraphrase, budget and scale prompts over all fourteen models, for a mid-market B2B company; rank is within the category. Comparisons are drawn for the top eight products in each category. Published under CC BY 4.0; the output is the models' output, and nothing here is a recommendation by the index.