EnactraAI introduced BuildingBench to assess whether coding agents can turn real-world images into coherent 3D buildings, and at what cost. It says GPT-6 Astra scored highest in its comparison at about 80% lower cost than Fable 5.1, while DeepSeek V4.1 Flash delivered “strong quality” at a median model cost of $2.31 per building. A user highlighting the benchmark says agents write Python code to create 3D meshes and textures without external 3D modeling software.