In September 2026, OpenAI released its new flagship model GPT-6 Astra, which on the 3D geometry-focused BenchCAD benchmark achieves direct reconstruction of editable CAD code (rather than static meshes) from multi-view photos, reaching a mean voxel IoU of 95.9% with tool configuration—a significant improvement over the previous generation GPT-5.6 Sol's 83.3% and also surpassing Anthropic Claude Fable 5.1's 84.3%. According to OpenAI, at equivalent scores, API costs are about 43% lower than Sol and about 86% lower than Fable 5.1. The key differentiator of BenchCAD lies in the output format: meshes are merely "shapes you can look at," whereas parametric CAD code can be edited, parameterized, and fed directly into standard design software. Astra is initially being rolled out to a limited set of institutions, with subsequent availability planned for ChatGPT Plus/Pro/Business/Enterprise, the OpenAI API, Microsoft Azure, and AWS Bedrock. For the additive manufacturing industry, this directly addresses the long-standing bottleneck of scan-to-CAD: scanner outputs—meshes or point clouds—cannot be used directly for CAD modeling, traditionally requiring manual redrawing or specialized reverse-engineering software. Take Backflip, founded by Markforged founder Greg Mark, as an example: its latest foundation model has reduced the digitization cost per part from about $1,500 to $10. Hardware manufacturer Creality has also partnered with KVS Ltd to launch an integrated scan-to-CAD workflow. Value for powder buyers: ① A sharp drop in part digitization costs will accelerate reverse engineering and replica design of legacy parts, potentially increasing the number of LPBF-printed parts and driving up demand for spherical metal powders; ② The dimensional accuracy of reverse-engineered parts is constrained by real scan data, so before parametric CAD output matures, the certification and batch consistency of high-precision powder grades (such as Ti-6Al-4V and 316L) will remain a core threshold for print shops when selecting powders; ③ Keep an eye on developments in the integration of domestic scanner manufacturers with AI design tools to capture order growth brought by cost reductions on the design side.