Benchmark
Blueprint Bench 2
An AI evaluation benchmark developed by Andon Labs that assesses spatial cognition and three-dimensional reconstruction capabilities. Models are required to infer relative positions, estimate dimensions, and generate scale-accurate floor plans from real indoor photographs, testing geometric reasoning, depth perception, and multi-view consistency.
Timeline (last 90 days)
Oct 5
Gemini 4 Argon ranks first on Blueprint Bench 2 (a benchmark for drawing floor plans from apartment photos), defeating Opus, Astra, Fable, and other competitors.
Unverified50%