Gemini Robotics 2.5 Launch - Close lid of black toolbox (Duo): leaderboard
Google DeepMind November 10, 2026 developer evaluation; Multi-step & contact-rich; task: Close lid of black toolbox; embodiment: Duo. Exact printed success-rate callout for Gemini Robotics 2 and Gemini Robotics 2.5 VLA models. Trial count, success rubric, held-out status, adaptation details and inference settings are not reported. No aggregate or calibration across the unlabeled tasks is implied. One of 16 labeled callouts from the wider 200-task evaluation, not a complete or random task sample. Duo denotes Franka Duo with two-fingered parallel grippers. Source: https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence/
Metric: Task success rate (%). Source: deepmind.google. Saturation forecast: Rough model projection: around 2026. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini Robotics 2.5 | 93 | |
| 2 | Gemini Robotics 2 | 0 |
Interactive version: theaggregate.ai/benchmark?slug=gemini-robotics-2-5-launch-close-lid-of-black-toolbox-duo · How It Works · Data refreshed daily, snapshot 2026-11-12.