As per the company blog post, the SKY-001 is a spatial world model designed to plan engineering and architectural designs directly in the visual world rather than relying solely on text or code. The model breaks complex tasks down using a two-stage architecture: The Hierarchical Composer and Construction World Model. \
Introducing SKY-001, the first Engineering World Model. On CAD and BIM tasks, it beats GPT-6 Astra and every frontier model we tested: -> 51-125x faster - > 150-210x cheaper -> Higher IOU than all frontier models SKY-001 lets you dream a physical object before you build it. 18 months of work. Built by the team out of Maluuba's deep learning lab (acq. by @MSRResearch). The first ‘AI General Engineer’ will be the backbone for Autonomous Manufacturing.
— Sam Pasupalak (@spisallyouneed) October 19, 2026
In the first stage, it designs the overall structure and breaks it down into smaller assemblies and individual parts. It then refines each component by filling in missing geometry, smoothing surfaces, and adding fine details. In the second stage, it determines how each component is built step by step.
Must read: 'AI models cannot be trusted': Microsoft warns of insider threats, expands Project Perception
In the X post, Sam Pasupalak has broken down how the SKY-001 AI model performs against other leading models. He claimed that SKY-001 is 51 to 125 times faster than the leading AI models his team tested.
The model is reportedly 150 to 210 times cheaper, depending on which AI model it is compared with. For a specific CAD task, SKY-001 reportedly costs just $0.005 per sample, while GPT-6 Astra costs $0.75. In terms of accuracy, SKY-001 scored 79.7% on the Intersection over Union (IoU) metric, compared with 79.1% for GPT-6 Astra, highlighting a difference of just 0.6 percentage points.
However, it should be noted that these results come from Skyfall AI's own tests and have not been independently verified.