I'm so sorry for your loss

2.5 Pro strikes again! Curious about the CV component as that's where I would have expected the highest lead (since the performance of 2.5Pro is substantially higher on traditional CV benchmarks).

Someone passed this wisdom to me today. Deep learning techniques working vs not working is two devils - your prior about the technique - your attention to details about implementation of the technique Need both to make it work.

Knowing the long line of amazing work from Sven's team, this is likely an amazing opportunity. Consider applying!

My daily AI tasks rely heavily on Flash (it excels in both speed+performance), and landing a major boost in it's performance is always a delight! Google DeepMind

There was a lot of announcements at IO, easy to overlook the new 2.5 Flash. It's pushing new boundaries in capability vs speed!

the biggest competition for VR is just R (reality) and when you're competing in an mature market you really need to make sure your product is 100x better in *some* way

FBGEMM github.com/pytorch/FBGEMM Can see fast implementations of QKV with rope and nope github.com/pytorch/FBGEMM…

Just so everyone knows, we have passed the point where you can tell what is AI at a glance (or even, in many cases, a close look) These were all made by me with text prompts alone using Veo 3.

We want Gemini to be able to think about problems on the frontier of math and science and ultimately assist the creation of new knowledge. These kinds of user experiences are very motivating to the team, and an exciting sign of what's to come!

The greatest feeling you can have as a scientist is realizing you’ve discovered something that no one else in the world knows yet When this happens, put your pencil down, take a deep breath and savor those three minutes before someone points out it was published fifty yrs ago

You go down a little only to come back stronger. Lots of progress ahead from us. Excited for the relentless model improvements coming.

We are hiring Research Scientists for our Machine Learning and Optimization team at Google DeepMind Bangalore. If you're passionate about cutting-edge AI research and building efficient, elastic, customized, and safe LLMs, we'd love to hear from you. We are looking for

Check out the 999 open models that Google has released on Hugging Face: huggingface.co/google (Comparative numbers: 387 for Microsoft, 33 for OpenAI, 0 for Anthropic).

For strong researchers, it doesn't get any better. You get to work with one of the best ML research teams in the world and contribute to Google's frontier models impacting billions of people!

Pre-training is not dead

Our report on 2.5 models is now publicly available. storage.googleapis.com/deepmind-media…

Say hello to the gemini-cli, a local CLI to help you build and maintain software with 1,000 free Gemini 2.5 Pro requests per day : )

better late than never lots of rough edges still but the team is grinding do share feedback github.com/google-gemini/…

AI researchers will literally negotiate $100 million comp packages by themselves but they won’t play poker for more than $50 buy-ins