AI

The best AI passes 37% of real engineering tasks. Most people would pass none

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/10/2026, 17:47:26 UTC.

Briefings

Xplore released the RWE-bench, a benchmark testing AI models on real-world mechanical, electrical, and aerospace engineering tasks. GPT-6 Astra achieved the highest score at 37%, significantly outperforming Opus at 29% and GPT-6.1 Sol at 25%. The study evaluated 12 distinct tasks, including diesel crankshaft design, FPGA memory routing, and CFD analysis, with no human scoring provided. Results indicate that while AI excels at code generation, human engineers struggle with complex physical simulations due to longer feedback loops, suggesting AI's advantage lies in speed and automated verification.

Key points

• GPT-6 Astra achieved the highest score of 37% on the RWE-bench benchmark.
• The benchmark evaluated 12 real-world engineering tasks including CAD design and CFD analysis.

Context

The RWE-bench tests AI capabilities against professional engineering tasks like crankshaft design and FPGA routing. This study highlights the gap between AI speed and human engineering experience, as humans often fail complex simulations despite strong coding skills.

Limitations

Human scores were not published by Xplore; only AI scores are available. The study covers 12 tasks but lacks data on human performance across all fields. Results are based on a single snapshot from October 2026 and do not account for long-term model evolution or human adaptation over time.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 09/11/2026, 04:06:42 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeiezm2mdlvd45nwr5g5tsfiisw44vuehauh2fokjbu7ulfgadehheu