OpenAI has announced progress in applying its generative and multimodal models to tasks in mechanical and electrical engineering, system design, simulation, and manufacturing preparation. The company says these capabilities are already enabling internal teams to test concepts and iterate more rapidly by automating portions of drawing, simulation setup, and cross‑discipline coordination.
Tasks and workflows targeted
Across projects, engineers must integrate mechanical designs, electronics, software, and manufacturing constraints using specialized tools. OpenAI’s models have been trained to assist with:
- creating and revising CAD models and manufacturing drawings,
- setting up and interpreting simulations,
- selecting electrical components and designing PCBs, including component placement and connections,
- generating UI/UX concepts, firmware, and software integrations.
The aim is to reduce time spent on repetitive manual work and free engineers to explore more design alternatives.
Internal results and metrics
OpenAI reports that improved frontier model capabilities over the past year have helped internal teams iterate faster. Within its hardware and infrastructure groups, usage of models measured in tokens has increased 23‑fold since February. Teams use Codex and ChatGPT to create and refine CAD models, prepare manufacturing drawings, and run design reviews across robotics and compute infrastructure projects.
BenchCAD 2.0, an internal benchmark, evaluates whether models can reconstruct 3D parts and assemblies and identify circuit board connections. A newer internal model achieved an average score of 0.875 on this benchmark, compared with 0.594 for GPT‑6 Astra at its highest reasoning setting. OpenAI presents these improvements as early evidence that AI can positively affect iteration speed in professional workflows, while noting the engineering lifecycle includes many other elements beyond design.
A practical test: building a device with GPT‑6 and Astra
In one experiment, OpenAI used GPT‑6 to define functional requirements and Astra to design an enclosure, the printed circuit board (PCB), and the necessary software integrations so users could interact with Codex. Astra derived the device’s aesthetic direction from OpenAI’s recent Work Louder collaboration and references from games and toys.
Two team members developed the test device with Astra in under a week as a side project, and a contract manufacturer produced a small batch of prototypes in under a month. Preparing the device for production still required several rounds of review: Astra did not consistently account for the manufacturer’s production limits, tolerances, or assembly details. OpenAI says these gaps highlight areas where future models need to improve.
Industry collaborations and field testing
OpenAI is working with a small set of industry collaborators to apply the models to real engineering problems. Partner engineers review model outputs for accuracy and manufacturability, helping OpenAI determine which tasks the models can support today and what improvements are necessary for safe, reliable use in industry.
The company is expanding collaboration with Chip Ganassi Racing (CGR) to accelerate engineering and prototyping. With CGR’s R&D team, OpenAI is testing workflows that convert 3D scans of existing parts into editable CAD models and 2D manufacturing drawings — a useful approach when original design files are unavailable. CGR engineers review the resulting models and drawings for accuracy and manufacturability, and their feedback guides further development.
OpenAI also announced a research collaboration with Mazda Motor Corporation. OpenAI’s research and Forward Deployed Engineering teams are working with Mazda’s aerodynamics and crash‑safety engineers to explore how models can help develop and evaluate new vehicle designs. The work uses simulations of airflow and structural behavior to search for performance improvements while considering styling and cost. OpenAI states that GPT‑6 Astra assists Mazda’s aerodynamics engineers by modeling new nose shapes, simulating airflow, and refining promising designs; agreement for use has been obtained from Mazda Motor Corporation.
Next steps and invitation to partners
OpenAI says early results are promising but that further development is required to achieve the reliability and production‑level detail needed for industry deployment. The company is inviting engineers, engineering firms, and software vendors involved in design, simulation, manufacturing, or production of complex physical products to collaborate on shaping training and development priorities. Interested parties are asked to fill out an online form to express interest and help identify high‑value use cases, tooling needs, and future opportunities in mechanical engineering, electrical engineering, electromechanical design, mechatronics, and related fields.
