What changed
On November 3, 2026, OpenAI announced that its models are being used to support engineering workflows, including design, simulation, and manufacturing. The company shared examples of how models like GPT-6 and Codex are being used to create devices, prepare drawings, and set up simulations. Astra, a model used in these efforts, was tested to create a device without direct human edits, involving mechanical, electrical, and software design.
What this means in practice
Engineers at OpenAI are using models to accelerate the design process, reducing manual work and improving iteration speeds. For instance, two team members designed a test device with Astra in under a week, and a contract manufacturer built prototypes in under a month. However, the process still required several rounds of review and iteration due to gaps in the model's understanding of production constraints.
The company also mentioned that its hardware and infrastructure teams have seen a 23-fold increase in model usage since February, with Codex and ChatGPT being used for tasks like creating CAD models and preparing manufacturing drawings.
OpenAI is collaborating with industry partners like Chip Ganassi Racing and Mazda to test these capabilities on real engineering problems, with feedback helping to identify areas for improvement.
Limitations and availability
While the models are already being used internally and in some collaborations, they are not yet fully integrated into all engineering workflows. The supplied announcement does not specify the exact availability of these models for external use. Additionally, the models still have limitations in accounting for production constraints and tolerances, which are being addressed through ongoing collaboration and testing.
