📢 First contact between a frontier model and robots!. Gemini Robotics is a SOTA generalist Vision-Language-Action model bringing frontier model intelligence to the physical world. It's an extremely capable model enabling dexterous, steerable, and general robot control. 🧵⬇️

Exciting work from AI2 on how 3D trajectories can act as a strong intermediate embodied reasoning representation for steering robot VLA policies!. And great to see SIMPLER continue to act as a useful benchmark for comparing VLAs trained on large-scale real world data 🚀

Robotics is not yet compute-bound! Mapping robotics to a predictive scaling law is critical if believe that the same trends in foundation models apply to embodied AI.

A few months ago, we created the Embodied Reasoning QA benchmark to measure foundational multimodal capabilities for understanding and acting in the real physical world. Happy to see this become an area of focus for the OpenAI team; congrats on the GPT-5 release! 🤖 🌎

We are living through the inflection point where general robotics evolves from cute academic sub-field to a ubiquitous technology. For those passionate about realizing this dream, consider joining my team at Google DeepMind!. If you believe in the trends that have powered.

everything is foundation model.

““In our main experiments, a "teacher" model with some trait T (such as liking owls or being misaligned) generates a dataset consisting solely of number sequences. Remarkably, a "student" model trained on this dataset learns T”. What?? 🤯.

Gemini Deep Think achieves gold-medal level results at the IMO 2025🥇based on official grading results. The unreasonable effectiveness of extended autoregressive “thinking” continues to push forward. Congrats to the whole team!

Unhacked benchmarks with no leakage are increasingly rare. Math and coding competitions are a great way to fairly evaluate reasoning models!. …all the top frontier models are already better than I ever will be on the hardest competition math problems 😅.

The most significant impact of superhuman AI may not be economic or technological, but spiritual. How do leaders of faith view AI today, and how will this relationship evolve as it reaches critical thresholds of intelligence and ubiquity?. On one hand, it's easy to frame strong.

This is the most polished short film powered by AI-generated music, images, and video that I have seen to date. Pretty astounding what systems like Veo and Sudo can do in the hands of talented creators 🤯 .

Link:

If you’re working on robotics and AI, the recent Stanford talk from @RussTedrake on scaling multitask robot manipulation is a mandatory watch, full stop. No marketing, no hype. Just solid hypothesis driven science, evidence backed claims. A gold mine in today’s landscape!

Despite the robot being in a totally alien landscape (nature! sunlight! crowds!), it kept chugging along. There were still performance challenges and (often hilarious) failures, but I’m so happy we got to show the world the potential of general robotic + AI methods like VLAs 🚀.

And that’s a wrap for #RSS2025 and the world’s first ever interactive live VLA robot demo 🤯. It felt like a magical early glimpse into the future — just talk to your robot, it will respond appropriately, and try to do something reasonable. New objects, new commands, new skills.

. Pierre Sermanet, Sean Kirmani, Stefani Karp, Stefano Saliceti, Steven Hansen, Sudeep Dasari, Ted Xiao, Thomas Lampe, Tianli Ding, Wenhao Yu, and Wentao Yuan! Thanks also to Xi Chen, Weicheng Kuo, and Paul Voigtlaende.

. Carolina Parada, David D'Ambrosio, Deepali Jain, Fei Xia, Jacky Liang, Jie Tan, Junkyung Kim, Kanishka Rao, Keerthana Gopalakrishnan, Ksenia Konyushkova, Laura Graessar, Lewis Chiang, Marissa Giustina, Mohit Sharma, Montserrat Gonzalez Arenas, Nicolas Heess, Peng Xu. .

Check out the detailed blogpost from @lgraesser3 and @xf1280: It covers important contributions from many on the Gemini Robotics Embodied Reasoning team: .Alex Hofer, Annie Xie, Arunkumar Byravan, Ashwin Balakrishna, Assaf Hurwitz Michaely. .