TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
OpenAI says its models are helping internal teams with design, simulation and manufacturing tasks, and it is testing engineering workflows with Chip Ganassi Racing and Mazda. The company reports faster internal iteration and improved scores on one engineering benchmark, while acknowledging that its models still miss production constraints and require human review.
OpenAI says its models are being used to support engineering work across mechanical design, electronics, software and manufacturing, and it has started or expanded collaborations with Mazda and Chip Ganassi Racing to test those capabilities on real projects. The company describes the work as early-stage: engineers review model outputs, and OpenAI says manufacturing constraints remain a source of errors.
In a report titled “Accelerating engineering for the physical world,” OpenAI described internal use of Codex and ChatGPT to create and revise computer-aided design models, prepare manufacturing drawings, set up simulations and review designs. The work spans its hardware and infrastructure teams, including robotics and computing infrastructure. OpenAI says usage across those teams, measured in tokens, has grown 23-fold since February. The report does not identify the year or provide an absolute usage figure.
The company also described a device-design test involving Astra. OpenAI says GPT-6 was used to define requirements and generate the device’s enclosure, printed circuit board, user interface, firmware and software integrations for connecting it to Codex. Two team members worked on the side project for under a week, and a contract manufacturer produced a small batch of prototypes in under a month. Those timing figures are OpenAI’s account of the project, not an independent comparison with conventional development.
OpenAI says the device still needed several rounds of review and iteration before it was ready for manufacturing. The model did not consistently account for the manufacturer’s production limits, tolerances or assembly details. The company is also working with Chip Ganassi Racing on converting 3D scans of parts into editable CAD models and manufacturing drawings, and with Mazda engineers on vehicle aerodynamics and crash-safety research. The partners’ engineers review outputs for accuracy and manufacturability.
Testing AI in Product Engineering
The announcement matters because engineering products involves more than generating a drawing or writing code: a design must fit physical components, satisfy performance goals and be practical to manufacture. OpenAI’s projects test whether models can coordinate work across design, electronics, software and simulation, potentially helping engineers explore more options or reduce some repetitive preparation work.
At this stage, the evidence described is preliminary. OpenAI says its teams are iterating faster, but it does not quantify the time saved or provide a controlled comparison. The collaborations could help assess how models perform against real production requirements, where mistakes can create costs or safety concerns. For readers in manufacturing and engineering, the central question is not simply whether a model can propose a design, but whether qualified engineers can verify and use its output reliably.
OpenAI reported that a newer internal model averaged 0.875 on BenchCAD 2.0, compared with 0.594 for GPT-6 Astra at its highest reasoning setting. The benchmark covers reconstruction of 3D parts and assemblies and identification of circuit-board connections. That result indicates progress on those tested tasks, but it does not establish performance across the full engineering lifecycle or prove that the model can independently deliver production-ready designs.
Products Worth Considering
Bose SoundLink Flex Bluetooth Speaker (2nd Gen) – Portable Outdoor Speaker with Hi‑Fi Audio, Waterproof and Dustproof, USB‑C, Up to 12 Hours
EcoFlow Delta 3 Portable Power Station – 1024 Wh, 1800 W
Logitech MX Keys Combo for Business Keyboard and Mouse – Graphite
As an affiliate, we earn on qualifying purchases.
From Internal Prototypes to Industry
OpenAI frames the work as part of a broader effort to apply its models to physical-product engineering. It says engineers must coordinate specialist tools and requirements across disciplines, and that its aim is to give them more time to explore and improve designs. The company says improved model capabilities over the past year have helped internal teams test ideas and iterate faster, while describing evidence of benefits in professional industry workflows as early.
The device project illustrates both the range and the limits of the approach. OpenAI says the model contributed to the device’s appearance, mechanical structure, circuit board and software. But the need for further manufacturing reviews shows that generating connected design elements is not the same as meeting a manufacturer’s detailed production requirements. The report does not say the prototype was shipped as a commercial product.
The industry work is also described as collaborative rather than autonomous. At Chip Ganassi Racing, engineers review CAD models and drawings derived from scans, especially in cases where original design files are unavailable. At Mazda, OpenAI teams are exploring design changes through airflow and structural simulations, with engineers involved in evaluation. OpenAI says it plans to share more about the Mazda work in the coming months.
Manufacturing Readiness Still Untested
The report does not provide independently verified measurements of productivity gains, error rates or costs. OpenAI’s claim that internal teams iterate faster is not accompanied by a quantified time comparison, and the reported 23-fold usage increase is measured in tokens, not completed engineering work. The starting and ending usage levels, and the year meant by “since February,” are not stated.
It is also unclear how often model-generated designs meet manufacturing requirements without substantial revisions, how the benchmark score relates to live engineering projects, or what safety and quality thresholds will be applied before broader deployment. The device project had multiple review rounds, and OpenAI explicitly identified gaps involving tolerances, production limits and assembly. The report does not give a deployment schedule or product availability for the tested capabilities.
Partner Tests and Results Ahead
OpenAI says it is working with a small group of industry collaborators, whose engineers review outputs and provide feedback on accuracy and practical use. The company is inviting engineers, engineering firms and software vendors involved in design, simulation, manufacturing or production to work with it on the effort.
For the Mazda collaboration, OpenAI says it expects to share more progress and results in the coming months. The work described focuses on using simulations to explore vehicle designs while accounting for aerodynamic performance, structural behavior, styling and cost. Further reporting will be needed to establish whether these experiments lead to measurable workflow improvements and how much human review remains necessary.
Key Questions
What engineering work is OpenAI testing its models on?
OpenAI describes support for CAD design, manufacturing drawings, simulations, circuit-board work and software integration. Its examples cover internal hardware teams and projects with Mazda and Chip Ganassi Racing.
Are the models producing manufacturing-ready products on their own?
OpenAI’s report does not establish that. The company says its device prototype needed several rounds of review, and that Astra did not consistently account for production limits, tolerances and assembly requirements.
What does the 23-fold usage figure mean?
OpenAI says token-measured model usage across its hardware teams grew 23-fold since February. The report does not name the year, provide absolute usage levels or show that completed engineering work rose by the same amount.
What are OpenAI and Mazda working on?
OpenAI says the collaboration involves Mazda’s aerodynamics and crash-safety engineers. One described test uses models to propose vehicle nose shapes and simulate airflow to explore ways of reducing drag; the company says more results are expected in the coming months.
Source: openai.com
