Google DeepMind releases Gemini Robotics 2, with only the reasoning model reachable in AI Studio

Sources: Google DeepMind, discussion
Summary: Google DeepMind announced Gemini Robotics 2. The release covers two vision-language-action models and an embodied reasoning model. The two vision-language-action models are limited to early-access partners. The embodied reasoning model is available in Google AI Studio.
Why it matters: The developer-facing surface of this release is planning and perception, not motor control.
Follow-up: Track whether the vision-language-action models reach general availability and on what terms.