Google DeepMind unveils Gemini Robotics 2.5 for robot dexterity
Google DeepMind has announced Gemini Robotics 2.5, which it describes as its most capable vision-language-action model for robots, with gains in dexterity, reliability and safety.
Key takeaways
- Google DeepMind announced Gemini Robotics 2.5 on 9 November 2026.
- The company describes it as its most capable vision-language-action model.
- DeepMind says the model improves robot dexterity, reliability and safety.
- Detailed performance figures and independent testing are not yet available.
What DeepMind announced
Google DeepMind announced Gemini Robotics 2.5 on 9 November 2026, according to the company's blog. The company calls it its most capable vision-language-action model, a type of AI model that lets robots see, understand language and act on both.
DeepMind says the new model enhances robot dexterity, reliability and safety. The announcement gives few further details, so independent testing and wider claims remain unconfirmed.
Why it matters
Vision-language-action models are a key step towards robots that can handle varied, real-world tasks rather than repeating fixed motions. If DeepMind's claims hold, Gemini Robotics 2.5 could make general-purpose robots more capable outside the lab.
Readers should treat the dexterity, reliability and safety improvements as the company's own statements until third-party evaluations appear.
Sources
- Gemini Robotics 2.5 accelerates real-world general dexterityGoogle DeepMind
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.
Related news
More AI news
Google DeepMind's AlphaProof Code verifies Signal encryption in Lean
Google DeepMind has announced AlphaProof Code, an AI system whose agents wrote three million lines of Lean proofs to verify Signal's core encryption.
Vercel data suggests AI models now peak within weeks of launch
Vercel's traffic data indicates AI models have a half-life of about six weeks, with most usage going to models less than two months old.
Adobe and Google expand partnership to bring Acrobat and creative tools into Gemini
Announced at Adobe MAX, the deal puts Adobe Acrobat and creative connectors inside Gemini, Gemini Enterprise and AI Mode in Search, while Google's AI models move into Adobe's apps.
OpenAI hires Odessia founder Francis Davidson in travel acqui-hire
OpenAI has acquired the year-old AI travel agent Odessia, and its founder, Sonder creator Francis Davidson, has joined the company. Terms were not disclosed and OpenAI has not announced any travel products.