GPT-6 Astra, an OpenAI AI, designs computer hardware; a human engineer reviews and revises the complex schematics.

Editorial illustration for OpenAI says GPT-6 Astra designed hardware in a week, but required revisions

OpenAI says GPT-6 Astra designed hardware in a week, but...

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

November 3, 2026 • 3 min read

OpenAI's hardware team built a gadget that lets people talk to the Codex coding agent, and they say GPT-6 Astra did nearly all the design work itself. The device started as a side project and came together in under a week, with Astra reportedly handling enclosure design, PCB layout, component selection, trace routing, firmware, and the user interface. A contract manufacturer turned out a small batch of prototypes within a month.

The company is framing this as proof that its models are closing in on real engineering competence, not just chat and code. OpenAI points to a new internal model that scores 0.875 on BenchCAD 2.0, a benchmark built to test whether a model can reconstruct 3D parts and assemblies and trace connections on circuit boards. GPT-6 Astra, by comparison, tops out at 0.594 on its highest reasoning setting. No details on the internal model have been released.

OpenAI says its hardware groups now lean on Codex and ChatGPT for CAD models, manufacturing drawings, and simulations in areas like robotics and compute infrastructure, with token usage up 23 times since February. The company is also said to be developing its own consumer hardware device, something between a smart speaker and a phone.

OpenAI's hardware team reportedly used GPT-6 Astra to design a device that lets users talk to the Codex coding agent, finishing it as a side project in under a week. According to OpenAI, Astra handled every step, from enclosure design and PCB layout with component selection and trace routing to firmware and the user interface.

OpenAI says GPT-6 Astra designed a hardware device from scratch in under a week - THE DECODER

Why this matters

The revision rounds are the real story here, not the one-week design sprint. OpenAI's own account admits Astra skipped manufacturing constraints, tolerances, and assembly details, the unglamorous stuff that separates a CAD file from a device a contract manufacturer can actually build at scale. That gap is exactly where hardware engineers earn their keep, and it's not closing on its own yet.

For founders eyeing AI-designed hardware as a shortcut around expensive design teams, the lesson is to budget for human review cycles regardless of how fast the initial draft comes together. For developers building on these models, the Codex-talking device itself is a more useful signal than the speed claim: OpenAI is clearly testing voice interfaces to its own coding agent, which hints at where product priorities sit. Researchers should note the company is leaning on a new "internal model" for its performance claims rather than giving outside labs anything to verify.

Impressive demo, real limits, and a reminder that OpenAI controls every number in this story.