MicroPlanBench: No SOTA VLM Beats 70% on Robot Next-Step Planning

A new benchmark pair, MicroPlanBench and MicroAnnotationBench, shows no SOTA multimodal model exceeds 70% on robot next-action planning, with about a third of predictions failing—GPT-6 Astra leads, while GPT-6.1 Sol matches closely at one-fifth the cost.

2026-10-20 ~ 2026-10-21 · 3 related posts

• Two New Robotics Benchmarks Released: MicroPlanBench and MicroAnnotationBench First Rankings — chris_j_paxton · 2026-10-20
• New egocentric benchmarks show SOTA VLMs fail ~1/3 of robot next-action predictions — mattbeane · 2026-10-20
• MicroPlanBench: no SOTA multimodal model tops 70% on next-subtask judgment for robots — animesh_garg · 2026-10-21