Astra scores the highest on Blueprint-bench2

Astra achieved the highest score on Blueprint-Bench 2, a test designed to evaluate AI agents' spatial reasoning. This benchmark requires agents to convert apartment photographs into accurate 2D floor plans. Each agent processes 50 apartments sequentially, analyzing approximately 20 interior photos per apartment to generate floor plans detailing room layouts, connections, and relative sizes. Agents utilize a persistent notepad for cross-apartment learning and iterative strategy refinement.