Release
Gemini Robotics 2.5 Put Humanoids on a Real Mercedes Line — 96% Success, and One Big Catch
DEV · By Paul Lovelace ·
The Brief
- Google DeepMind announced Gemini Robotics 2.5 on November 10, 2026, describing it as its most capable vision-language-action (VLA) model to date.
- Apptronik's Apollo 2 humanoid, running the model, picked and packed automotive parts in a Mercedes line test.
- The headline reports a 96% success rate on these tasks; the source excerpt is truncated and does not give full details.
Why it matters
It is a reported real-world test of a VLA model controlling a humanoid robot in an automotive production setting, so developers working on robot control may want to follow the results. The source's stated caveat ("one big catch") is cut off in the available text.
Who’s affected
- Robotics developers
- Teams evaluating VLA models for humanoid robots
Yesterday, November 10, 2026, Google DeepMind announced Gemini Robotics 2.5 — its most capable vision-language-action (VLA) model to date. The headline number is a good one: Apptronik's Apollo 2 humanoid, running this model, picked and packed automotive parts on active…
Timeline
Google DeepMind announces Gemini Robotics 2.5
DEV Community: ai
Gemini Robotics 2.5 Put Humanoids on a Real Mercedes Line — 96% Success, and One Big Catch