OpenAI Accelerates Physical Product Engineering With AI Models

OpenAI is outlining how its frontier models are being extended to support mechanical and electrical engineering, with internal teams and external partners using the technology to accelerate design, simulation, prototyping, and manufacturing of physical products. The company said its models are being taught to handle workflows that traditionally require specialized tools, including preparing drawings, setting up simulations, and coordinating changes across mechanical, electronic, software, and production disciplines. Over the past year, improvements in model capabilities have helped OpenAI hardware teams test ideas faster, while early evidence suggests similar gains may reach professional industrial workflows. A key internal demonstration involved using GPT-6 Astra to define functional requirements and design a voice-enabled device for interacting with Codex, including the enclosure, printed circuit board, and software integrations, without direct human edits. Two team members completed the concept in less than a week as a side project, and a contract manufacturer produced a small batch of prototypes in under a month. OpenAI noted that the prototype still required several rounds of review because the model did not consistently account for manufacturing limits, tolerances, and assembly details. Those shortcomings are being used to identify requirements for future models. OpenAI said its hardware and infrastructure teams are using Codex and ChatGPT to create and revise computer-aided design models, prepare manufacturing drawings, configure simulations, and review designs for robotics and compute infrastructure. Since February, token usage across hardware teams has increased 23-fold. The company is also tracking engineering-specific capabilities through BenchCAD 2.0, which tests whether models can reconstruct 3D parts and assemblies and identify circuit-board connections. In that benchmark, a newer internal model scored an average of 0.875, compared with 0.594 for GPT-6 Astra at its highest reasoning setting. Beyond internal use, OpenAI is working with industry collaborators on real engineering problems. The company expanded its collaboration with Chip Ganassi Racing, where researchers and the team’s R&D group are testing workflows that convert 3D scans of existing parts into editable CAD models and 2D manufacturing drawings, especially when original design files are unavailable. Engineers review the outputs for accuracy and manufacturability, helping identify tasks where AI can reduce manual work and where further improvements are needed. OpenAI also announced a research collaboration with Mazda. Its research and Forward Deployed Engineering teams are working with Mazda aerodynamics and crash-safety engineers to explore how model-driven simulations can evaluate vehicle designs, test airflow and structural behavior, and balance performance, styling, and cost. The company plans to share progress in the coming months. OpenAI said it is seeking additional collaborators from mechanical engineering, electrical engineering, electromechanical design, mechatronics, manufacturing, and related fields to help shape model training, identify high-value use cases, and improve software tooling. The initiative is positioned as an early effort to expand AI agents beyond digital workflows and into safe, reliable support for designing and building physical products.