GPT-6 Astra Designed Hardware From Scratch — But It Needed Fixes

According to The Decoder, OpenAI's hardware team recently used GPT-6 Astra to design a physical device that lets users talk to the Codex coding agent. The entire project was pulled off as a side task in under a week.

How Astra Handled the Design Process

The model took charge of every step in the pipeline. This included enclosure design, printed circuit board layout, component selection, trace routing, firmware, and the user interface.

A contract manufacturer managed to produce a small batch of prototypes in under a month. However, the process wasn't frictionless. Astra failed to account for real-world manufacturing constraints, assembly details, and physical tolerances, which forced several revision rounds before production could move forward.

The Performance Leap in Engineering Benchmarks

OpenAI points to a massive jump in reasoning capabilities. While GPT-6 Astra hits a score of 0.594 at its highest reasoning level on the BenchCAD 2.0 benchmark, a newer unreleased internal model reaches 0.875.

This benchmark tests whether language models can reconstruct 3D parts, manage complex assemblies, and accurately identify electronic connections on PCBs. Meanwhile, internal token usage for hardware and robotics projects has surged 23x since February.

Is This Ready for Production Teams?

Not yet. While AI can draft enclosures and route circuit boards in days, the need for human oversight on physical tolerances remains absolute. If you plan to test generative AI for physical prototyping, budget extra time for mechanical revisions.