Osegi
AI·1 min read

Google DeepMind rolls out Gemini Robotics 2.5 to partners

Google DeepMind has begun rolling out Gemini Robotics 2.5 to partners, saying it improves task success, speed and physical safety controls over earlier baselines.

A robotic arm performing a task in a lab, illustrating Google DeepMind's Gemini Robotics 2.5 model.
Image: techinasia.com

Google DeepMind is rolling out Gemini Robotics 2.5 to partners, the company said, presenting the model as an improvement over earlier baselines in three areas: task success rates, completion speed and physical safety controls.

The release is aimed at partners rather than the general public. Google did not provide further detail in the material available, so specifics such as benchmark figures, the list of partners or supported robot hardware are not confirmed here.

Why it matters

Gemini Robotics is Google's effort to bring its Gemini AI models into the physical world, where software has to control machines that act on real objects and people. In that setting, reliability and safety matter as much as raw capability. A robot that completes tasks more often and more quickly is more useful, but one that respects physical safety limits is far easier to deploy around humans.

By highlighting safety controls alongside success rates and speed, Google is signalling that it sees these as linked requirements for commercial robotics, not separate goals.

What comes next

The partner rollout will give outside developers and robot makers a chance to test the model on their own systems and tasks. How widely Gemini Robotics 2.5 becomes available, and how its gains hold up beyond Google's own comparisons, will be the next questions to answer.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:google deepmindgemini roboticsroboticsai modelsphysical ai

More from this day

See all →

More in: AI

See all →