OpenAI Accelerates Physical Engineering With GPT-6 Astra and Codex
OpenAI published an update on November 3, 2026 describing how its models, including GPT-6 Astra and Codex, help engineers design, simulate, and build physical products. The post's main evidence is a test device designed with GPT-6 Astra, a 23-fold increase in model usage across OpenAI's hardware teams since February, and new or expanded engineering collaborations with Chip Ganassi Racing and Mazda.
What OpenAI announced
OpenAI is sharing advances in how its models help engineers design, simulate, and build physical products. The work spans mechanical engineering, electrical engineering, and end-to-end system design. OpenAI says these capabilities are already helping internal teams test ideas and iterate faster.
OpenAI frames the problem around tool fragmentation. Engineers must combine mechanical designs, electronics, software, and manufacturing requirements using specialized tools. OpenAI is teaching models to support that work, from preparing drawings and setting up simulations to coordinating changes across disciplines. The stated goal is to give engineers more time to explore ideas and improve designs.
OpenAI says improved frontier model capabilities over the past year have helped internal teams test new ideas and iterate faster, and it sees early evidence of these capabilities improving professional workflows in industry. OpenAI is working with collaborators to find where the capabilities make the biggest difference and what needs to improve. OpenAI invited engineers, engineering firms, and software vendors involved in design, simulation, manufacturing, or production of complex physical products to work with OpenAI through its collaboration form.
A test device designed with GPT-6 Astra
OpenAI tested Astra's state-of-the-art performance in multimodal and computer-use workflows by seeing whether a device could be created without any direct human edits. The team used GPT-6 to define functional requirements and to design the enclosure, printed circuit board, and software integrations for a device that users can use to talk to Codex.
Astra produced work across four areas:
- Aesthetics and design: Astra developed the device's look, drawing on OpenAI's recent Work Louder collaboration and references from games and toys.
- Mechanical design: Astra designed the enclosure and the internal structures that hold components in place.
- Electrical design: Astra selected electrical components and designed the PCB, including component placement and connections, to fit within the enclosure.
- UI/UX and software: Astra designed the user interface and generated the firmware and software integrations needed to connect the device to Codex.
Two team members designed the test device with Astra in under a week while working on it as a side project. A contract manufacturer built a small batch of prototypes in under a month.
Manufacturing limits remained a gap
OpenAI states that getting the test device ready for manufacturing still required several rounds of review and iteration. Astra did not consistently account for the manufacturer's production limits, tolerances, or details required for assembly. OpenAI says these gaps are helping it identify what future models need to do better.
The qualification matters: the demonstration was a prototype workflow, not a fully autonomous path from concept to production. Human and manufacturing review remained necessary.
OpenAI hardware teams and the 23-fold usage increase
OpenAI's hardware and infrastructure teams use Codex and ChatGPT to create and revise computer-aided design models, prepare manufacturing drawings, set up simulations, and review designs. This work spans robotics and compute infrastructure. Since February, model usage across OpenAI's hardware teams, measured in tokens, has grown 23-fold.
That number is a usage metric, not a productivity benchmark. OpenAI presents it as evidence of adoption inside its own hardware work, not as proof that every engineering task is faster.
BenchCAD 2.0 results
OpenAI is measuring progress on tasks that combine coding with engineering knowledge. BenchCAD 2.0 tests whether models can reconstruct 3D parts and assemblies and identify connections in circuit boards. A newer internal model achieved an average score of 0.875 on BenchCAD 2.0, compared with 0.594 for GPT-6 Astra at its highest reasoning setting.
OpenAI notes that the engineering lifecycle includes many other elements beyond design. OpenAI views improvements on this benchmark as early evidence that AI models can positively affect iteration speed in professional workflows.
The difference between the internal model and GPT-6 Astra is substantial on this benchmark. It suggests that engineering-specific capability is still improving and that the benchmark is not yet a complete measure of engineering competence.
Industry collaborations: Chip Ganassi Racing and Mazda
OpenAI is expanding its collaboration with Chip Ganassi Racing to explore how OpenAI models can accelerate engineering and prototyping. Working with Chip Ganassi Racing's R&D team, OpenAI is testing workflows that turn 3D scans of existing parts into editable CAD models and 2D manufacturing drawings. This is particularly useful when the original design files are unavailable.
Chip Ganassi Racing engineers review the models and drawings for accuracy and manufacturability. Their feedback helps OpenAI identify where models can reduce manual work and where further improvements are needed.
OpenAI also announced a new research collaboration with Mazda. OpenAI's research and Forward Deployed Engineering teams are working with Mazda's aerodynamics and crash-safety engineers to explore how OpenAI models can help develop and evaluate new vehicle designs.
The Mazda work uses simulations of airflow and structural behavior to explore improvements in performance while accounting for styling and cost. OpenAI says it will share more progress and results in the coming months.
In the aerodynamics work described by OpenAI, GPT-6 Astra helps Mazda's aerodynamics engineers explore ways to reduce vehicle drag by modeling new nose shapes, simulating airflow, and refining the most promising designs.
Human review and safe deployment
OpenAI is working with a small group of industry collaborators to test these capabilities on real engineering problems. Their engineers review model outputs for accuracy and practical use. That review helps OpenAI understand which tasks models can support today and what needs to improve for safe and reliable use.
The collaborations place human engineers in a review role. Chip Ganassi Racing engineers check models and drawings for accuracy and manufacturability. Mazda engineers work with OpenAI on vehicle design evaluation. OpenAI describes this structure as a way to understand both useful tasks and remaining gaps.
How engineers can collaborate
OpenAI says it has been working closely with industry partners and early adopters to push the bounds of what AI agents can do in hardware and manufacturing. OpenAI says early results show promise and it is looking for collaborators to accelerate capability improvements and ensure safe deployment into industry.
OpenAI invites people interested in shaping the training and development of its frontier models by identifying high-value use cases, software tooling, or future opportunities in mechanical engineering, electrical engineering, electromechanical design, mechatronics, or a related field to fill out its collaboration form.
What this means
OpenAI's post positions AI engineering agents as assistants inside existing engineering workflows, not replacements for engineering review. The strongest evidence in the post is the test device designed with GPT-6 Astra in under a week and the 23-fold growth in token usage across OpenAI's hardware teams. The clearest limits are also stated: Astra did not consistently account for production limits, tolerances, and assembly details, and manufacturing still required several rounds of review and iteration.
The BenchCAD 2.0 results show measurable progress on design-adjacent tasks, but OpenAI itself frames those results as early evidence rather than a complete measure of engineering capability. The Chip Ganassi Racing and Mazda collaborations are designed to test real-world usefulness and safety with practicing engineers reviewing outputs.