Open Sourced SML
A standardized, multidimensional Semantic Modeling Language to fully express the most sophisticated business concepts
Developer Hub
The SML specification, a library of models you can clone and run today, converters for models you’ve already built, and the API reference for taking one to production. All open-source on GitHub under Apache-2.0.
Free courses in AtScale Academy cover the concepts. AtScale is also a member of Apache Ossie, the open standard for moving semantic models between analytics, BI and AI platforms, now an Apache Incubator project.
Pre-Built Model Library
Domain-specific semantic models ready to use with any BI tool
Product Documentation
Explore detailed how-tos, installation guides, and the SML specification
Knowledge Base
Find documentation, guides, and resources to help you get the most out of AtScale
Connect AI to your semantic layer
Point an agent or MCP client at governed models so it queries defined metrics instead of guessing from raw schema.
API Reference
Query models, manage catalogs and repositories, control aggregate builds, and connect over XMLA. Full endpoint reference for the AtScale container API.
Apache OSSIE
The open standard for exchanging semantic models across analytics, BI and AI platforms, now an Apache Incubator project. AtScale is a member.
AtScale Academy
Free, self-paced courses on the semantic layer and the platform. Start with Foundations, then build your first governed model in Modeling 101.
Tutorials
Step-by-step guides to semantic modeling concepts. Learn the ideas, then build them on your own data.
Benchmarking the Semantic Layer for AI
BIRD-Interact is an independent, peer-reviewed benchmark that tests how often AI answers real business questions correctly. 600 tasks, 22 messy PostgreSQL databases, built by academic researchers and Google Cloud.
- Without a semantic layer, AI gets about one enterprise question in five right (21.2% average, September 2026)
- Adding AtScale’s AI-generated semantic layer lifted first-question accuracy by 72% to 137% across seven models, reaching up to 62.2% on the 410 business questions (best of three runs; 55% to 108% on the cold first run; preliminary).
- Every model improved, open-weight and frontier alike. Even the weakest model with AtScale (44.6%) beat the strongest model on its own (36.1%).
- Anyone can rerun it. The harness, tasks, databases, and scoring code are all public.Learn from real-world strategies and case studies shared by practitioners
What is Semantic Modeling Language?
Semantic Modeling Language (SML) is an open-source, YAML-based standard for defining semantic models that work across BI and AI tools.
Why it matters:
- Define metrics and dimensions once, reuse them everywhere
- Human-readable, code-friendly, and version-controlled in Git
- Supports complex analytics (hierarchies, relationships, custom measures) Open and extensible for any business logic
Semantic Layer Summit, Presented by AtScale
The premier virtual event for data leaders, engineers, and analysts exploring the semantic layer’s role in modern analytics and AI.
- Learn from real-world strategies and case studies shared by practitioners
- Stay ahead of emerging trends in metrics, data modeling, and AI-ready infrastructure
- Connect with a community of experts shaping the future of consistent, scalable data
See AtScale In Action
Watch AI agents get the right answer from a semantic layer, through MCP
Watch a Demo Video