OpenAI says GPT-⁠6 Astra designed a device that talks to Codex, with no direct human edits

A test device designed by Astra

In a Nov. 3 post titled “Accelerating engineering for the physical world,” OpenAI described how its models are being used to design physical products. The company tested GPT-6 Astra’s multimodal and computer use capabilities by seeing whether it could create a device without any direct human edits. The device lets users talk to Codex. According to OpenAI, GPT-6 defined the functional requirements. Astra developed the device’s look, drawing on the company’s Work Louder collaboration and references from games and toys. The model also designed the enclosure and the internal structures that hold the components in place. It selected the electrical components and laid out the printed circuit board (PCB), including placement and connections. Astra also produced the interface, the firmware and the software connecting the device to Codex.

Two team members designed the device with Astra in under a week as a side project. A contract manufacturer built a small batch of prototypes in under a month. Getting it ready for manufacturing still took several rounds of review and iteration. According to OpenAI, Astra did not consistently account for the manufacturer’s production limits, tolerances or assembly details. The company says these gaps show what future models need to do better. The device is a test prototype, not a product for sale.

OpenAI@OpenAI

ChatGPT and Codex are helping engineers build for the physical world.

We’re sharing our progress across mechanical design, electrical engineering, and simulation—from helping our hardware teams iterate faster to testing new workflows with industry collaborators.

Usage up 23x on hardware teams

OpenAI’s hardware and infrastructure teams use Codex and ChatGPT to create CAD models, prepare manufacturing drawings, set up simulations and review designs. The work spans robotics and compute infrastructure. The company says model usage across these teams, measured in tokens, has grown 23-fold since February.

OpenAI also tracks progress with a test called BenchCAD 2.0. It checks whether models can reconstruct 3D parts and assemblies and identify connections in circuit boards. By OpenAI’s own measurement, a newer internal model scored 0.875 on average, compared with 0.594 for GPT-6 Astra at its highest reasoning setting.

Chip Ganassi Racing and Mazda

OpenAI is testing these capabilities on real engineering problems with a small group of industry collaborators, whose engineers review the outputs for accuracy and practical use. It is expanding its work with the racing team Chip Ganassi Racing. There, models turn 3D scans of existing parts into editable CAD models and 2D manufacturing drawings, which helps when the original design files are unavailable.

OpenAI also announced a new research collaboration with Mazda. Its teams are working with Mazda’s aerodynamics and crash-safety engineers on simulations of airflow and structural behavior. According to Akihiro Nakata, a principal engineer at Mazda, Astra helps the team see which features matter for aerodynamics by simulating dozens of variations. OpenAI says it will share results in the coming months.