Google Opens 85-Task Engineering Benchmark and Agent Competition

Google AI & Infrastructure released 29 public tasks and evaluation tools from the Physical-Engineering’s Last Exam (PLE), an 85-task benchmark for engineering AI agents, and launched an open competition with a leaderboard. The benchmark covers mechanical, electrical, civil, and structural engineering tasks requiring agents to produce professional deliverables meeting engineering codes and constraints. Tasks run in containerized environments with real-world project files, graded by expert-authors