AtScale says semantic layer improved all seven AI models tested
The company reported higher accuracy on business-data questions for seven large language models when used with a semantic layer.
Event / data period: 2026-10-15
Published: · Updated:
On 15 October, AtScale published results from an evaluation of seven large language models on the BIRD-Interact benchmark. The company said every tested model was more accurate on the 410 business-data questions when paired with a semantic layer, and that an open-weight model reached a level close to frontier models.
The findings are from AtScale’s own experiment; the company labels them preliminary and says the BIRD team has not verified them. The test examines the difference between models querying raw enterprise tables and querying through defined business metrics. It does not establish that every company will see the same gains.
Data
AtScale says semantic layer improved all seven AI models tested
AtScale BIRD-Interact evaluation
- · Occurred
Evaluation scope
Seven models and 410 business-data questions; improvement with the semantic layer was reported for each model.
Methodology
Figures reflect the company’s preliminary evaluation; they are not verified by BIRD and are not independent results.
Updated:
Subjects and places
Subject proposals
- Artificial intelligenceSuggested
- Subject suggested from the title or summary: AIwww.atscale.com ↗
Geographic references
Only source-provided geographic references are shown. Opening an external map is optional. Optional story maps use OpenStreetMap. Opening a map sends your IP address and this site’s address to the map provider; location data remains readable as text without loading the map.
- United States · Place role: Coverage · Place type: Country or area · Location precision: RegionalEvidence
- Explicit preserved editorial country/region; no city or national centroid inferredwww.atscale.com ↗