Semantic Modeling Language (SML)

Semantic Modeling Language (SML) is an open-source, YAML-based standard for defining semantic models that work across BI and AI tools. Developed and open-sourced by AtScale in 2024, SML targets data engineers, analysts, and business users who need a single, portable definition of metrics and dimensions to eliminate inconsistency across tools. It is designed for organizations that have outgrown siloed metric definitions in individual BI platforms and want a governed, reusable semantic layer that can feed any downstream tool — from Tableau and Power BI to AI agents. SML is particularly suited for enterprises in finance, healthcare, retail, manufacturing, and CPG, where complex multi-dimensional analytics (hierarchies, semi-additive measures, many-to-many relationships) are common.

SML is object-oriented, promoting composability and inheritance so that semantic objects can be shared within and across organizations. It supports complex analytics including hierarchies, relationships, custom measures, and cell-based expressions. The language is built on YAML, making it human-readable, code-friendly, and fully compatible with Git for version control and CI/CD pipelines. SML is extensible — its syntax can be enhanced to support additional properties — and is Apache-licensed, free to use in any application. AtScale has also released pre-built semantic models for standard datasets like TPC-DS, Worldwide Importers, AdventureWorks, Snowplow, and CRISP, with plans to add models for SaaS applications such as Salesforce, Google Analytics, and Jira. Future releases will include helper classes for programmatic read/write and semantic translators to migrate from other languages like dbt Labs' semantic layer.

SML competes directly with proprietary semantic modeling approaches from vendors such as dbt Labs (dbt Semantic Layer), Looker (LookML), Cube.dev, Denodo, and Strategy Mosaic. Unlike LookML, which is tied to Looker's platform, SML is vendor-neutral and open-source, aiming to become an industry standard. AtScale's own platform implements SML, but the specification is designed to be portable across any semantic layer engine. Strategy Mosaic positions itself as an enterprise alternative with AI-powered no-code modeling and continuous governance, while dbt's semantic layer is tightly integrated with its transformation workflow. SML's open-source nature and multidimensional depth (covering semi-additive measures, many-to-many relationships, and cell-based expressions) differentiate it from simpler tabular-focused competitors.

The primary trade-off with SML is its relative newness — the specification was open-sourced in late 2024, so the ecosystem of tools, translators, and community models is still maturing. Organizations adopting SML today may need to build custom integrations or wait for third-party support. While SML is comprehensive, its YAML syntax can become verbose for very large models with hundreds of objects. There is no explicit cons documented in the sources, but potential adopters should consider the learning curve for teams unfamiliar with YAML-based modeling and the dependency on AtScale's ongoing stewardship for specification evolution. Additionally, SML currently lacks built-in governance features (e.g., automated lineage, policy enforcement) that enterprise platforms like Strategy Mosaic provide out of the box.

Free and open-source under the Apache 2.0 license. No licensing costs for using the SML specification. AtScale's commercial semantic layer platform (which implements SML) requires a separate paid subscription.