OpenAI unveils a more powerful engineering model: CAD evaluation score of 0.875, significantly surpassing Astra.

Beating AI News Flash: OpenAI has revealed an unreleased internal model. In the BenchCAD 2.0 engineering design evaluation, it achieved an average score of 0.875, significantly surpassing GPT-6 Astra's highest reasoning tier score of 0.594, a relative improvement of approximately 47%. The model name and release date have not yet been announced.

BenchCAD 2.0 primarily tests whether AI can reconstruct 3D parts from drawings, assemble mechanical structures, and identify circuit connections. Astra previously ranked first in public evaluations but also made elementary mistakes: it identified the 3.3V, 5V, and ground lines on a circuit board as the same net. If such errors enter actual circuit design, they could directly cause hardware failures.

OpenAI also demonstrated Astra's machine-building experiment. Two employees spent less than a week designing a physical device capable of interacting with Codex via voice. Astra completed the enclosure, internal structure, circuit board, and software design. However, actual production still requires multiple rounds of manual review and adjustment. The contract manufacturer took nearly a month to produce the first batch of prototypes.

OpenAI is also collaborating with racing team Chip Ganassi Racing to convert 3D scans of parts into editable CAD models and manufacturing drawings; and with Mazda to explore automotive designs that reduce drag by adjusting front-end styling and simulating airflow.