New Embodied Benchmark Leaves Best Agents at 32.9%

A multi-university team including Northwestern, CMU, Stanford and UC Berkeley released Embodied First Exam, a benchmark of hundreds of embodied coding tasks. The best agent solved only 32.9% of tasks, highlighting a major gap in embodied AI capabilities.

2026-11-09 ~ 2026-11-09 · 2 related posts

• Embodied First Exam: top agent GPT-6 Astra solves just 32.9% of 456 core robotics tasks — ManlingLi_ · 2026-11-09
• Embodied First Exam: 1,090 embodied tasks stump coding agents, best solves just 32.9% — LINJIEFUN · 2026-11-09