DBRX — benchmark results

Databricks' fine-grained MoE model (132B total/36B active, 16 experts pick 4) trained on 12T tokens, released March 2024 under the Databricks Open Model License. Provider: Databricks. Released 2024-03-27. Access: Open.

Unified ELO 1419 ± 32, rank #1136 of 1776 rated models, from 29 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open CoT - LSAT Logical Reasoning20.59CoT Gain (%)95
Open CoT - LSAT Reading Comprehension21.19CoT Gain (%)90.1
URIAL-Bench - Math5.9Judge Score (0-10)88.9
URIAL-Bench - Overall7.22Judge Score (0-10)88.9
URIAL-Bench - Roleplay7.93Judge Score (0-10)88.9
URIAL-Bench - Turn 18.04Judge Score (0-10)88.9
Open LLM Leaderboard - GPQA12.19Score84.9
URIAL-Bench - Coding4.7Judge Score (0-10)83.3
URIAL-Bench - Humanities9.32Judge Score (0-10)83.3
URIAL-Bench - Reasoning5.9Judge Score (0-10)83.3
URIAL-Bench - STEM8.43Judge Score (0-10)83.3
URIAL-Bench - Turn 26.4Judge Score (0-10)83.3

Interactive version: theaggregate.ai/model?slug=dbrx · How the rankings work · Data refreshed daily, snapshot 2026-07-22.