Global evaluation · 75 cases
Evaluation scoreboard
Multi-step spatial reasoning results across 27 frontier LLMs. Expand any family to inspect the Coverage, Geometry and Semantic breakdown for every difficulty level.
27
Frontier models
25
Test families
75
Test cases
95%
Best overall score
|
Loading evaluation data… |
Strong (> 80)
Partial (40–80)
Weak (< 40)
Scores are deterministic and rendered from the parametric benchmark database.