Issue 1, Topic 1
AI Model Benchmark for AEC
Last updated: September 22, 2026
8 AI models tested head-to-head on 5 real architectural tasks, every model on its free default tier. ChatGPT won the Ultimate Ranking (71.4%): #1 in Tasks 4 and 5, and the only model to pass the floor plan task.
The 5 tasks tested
- Conceptual facade image generation, with specific design constraints
- Apartment floor plan: reasoning, then HD image generation
- A working GHPython script for Grasshopper (Rhino 8)
- Parametric logic for a rhomboidal panel facade
- A real building-envelope thermal transmittance requirement
The Ultimate Ranking
1. ChatGPT (71.4%) · Places 2 to 8 are in the full report
ChatGPT took first by ranking #1 in Tasks 4 and 5 and being the only model to pass the floor plan task. Nobody won every task.
The most interesting finding: the best-documented code did not win. ChatGPT wrote the best-documented Grasshopper script of the group, yet placed the camera 100,000 units above the point the prompt specified, and ranked only #4 on that task. Polish is not correctness.
What you get
The full interactive report (every model, every task, every image side by side) and a print-ready PDF, both delivered instantly after purchase.
