OpenAI published results on November 3, 2026 describing how its frontier models are being applied to engineering tasks that span mechanical design, electronics, software and manufacturing. The company says models are being taught to prepare drawings, set up simulations and coordinate cross-discipline changes to help engineers iterate faster.
Device design and early limits
In one internal experiment, GPT‑6 Astra was used to define functional requirements and produce an enclosure, printed circuit board (PCB) layout and the software integrations for a device that communicates with Codex. Two team members worked on the project as a side effort and completed the design phase in under a week; a contract manufacturer produced a small batch of prototypes in under a month. OpenAI notes that preparing the device for full manufacturing still required several review and iteration rounds, and that Astra did not consistently account for a manufacturer’s production limits, tolerances or assembly details.
Internal use and benchmarks
Hardware and infrastructure teams at OpenAI report using Codex and ChatGPT to create and revise CAD models, convert 3D designs into production-ready 2D drawings, set up simulations and review designs for robotics and compute infrastructure. The company reports that model usage across its hardware teams, measured in tokens, has increased 23-fold since February.
BenchCAD 2.0, an internal benchmark, tests whether models can reconstruct 3D parts and assemblies and identify circuit-board connections. OpenAI reported an average score of 0.875 for a newer internal model compared with 0.594 for GPT‑6 Astra at its highest reasoning setting.
“Codex has dramatically accelerated how we move from an idea to an engineered part. We’re iterating faster across CAD, converting 3D designs into production-ready 2D drawings, and setting up simulations and FEA with far less manual work,” said Grace Lam, Forward Deployed Engineer at OpenAI.
Industry collaborations
OpenAI said it is testing these capabilities with a small group of industry partners. Work with Chip Ganassi Racing (CGR) focuses on converting 3D scans of existing parts into editable CAD models and 2D manufacturing drawings when original files are unavailable. “The model has been useful for going from raw scan data to a solid body. I have been working with Codex directly in CAD…this has become a faster, more hands-off process for me,” said Matthew Ortman, R&D Engineer at Chip Ganassi Racing.
A research collaboration with Mazda pairs OpenAI teams with Mazda’s aerodynamics and crash-safety engineers to model airflow and structural behavior. OpenAI says GPT‑6 Astra helped explore nose-shape variations to reduce drag while accounting for styling and cost. “By exploring and simulating dozens of variations, Astra in Codex helps us understand which features matter for aerodynamics and where designers have room to explore,” said Akihiro Nakata, Principal Engineer at Mazda.
OpenAI stated it will continue developing these capabilities, share further results in coming months, and is seeking engineers, engineering firms and software vendors involved in mechanical, electrical or electromechanical design to collaborate on further testing and deployment.
Original source: OpenAI News