GPT-6.2 Astra tops Blueprint-Bench 2 spatial reasoning benchmark, beating human baseline

141_1337 · reddit · 2026-11-20

Per an Andon Labs post, GPT-6.2 Astra ranks #1 on Blueprint-Bench 2, a benchmark where AI agents reconstruct apartment floor plans from interior photos alone. Astra outscores every other model tested and edges past the 0.59 human baseline; GPT-6.2 Sol ranks #3, while Claude, Gemini, Grok, Kimi and others all fall below the human line.

Related event: GPT-6.2 Astra Tops Blueprint-Bench 2, Beating Human Baseline in 3D Understanding(3 posts)→

Original post →

More from Models

Models channel →