Gemma 3 is out!
We are focused on bringing you open models with best capabilities while being fast and easy to deploy:
- 27B lands an ELO of 1338, all the while still fitting on 1 single H100!
- vision support to process mixed image/video/text content
- extended context window
I'm thrilled to announce I have joined @DeepMind to help their extraordinary team continue to lead in their quest to build AGI. More to come. I'm also so happy to reconnect with so many friends I worked with a decade ago. Let the adventure begin!
Today I am happy to share that Google AI Studio and the Gemini Developer API (along with our teams) are moving over to @GoogleDeepMind!
This move will allow us to double down on our already deep collaboration and accelerate the research to developer pipeline. Time to build!
Gemma is expanding.... we just announced CodeGemma, a version of Gemma tuned for code generation. And bonus... Gemma is now bumped to v1.1, addressing lots of feedback we got.
Congrats Gemma team for one more amazing release!
For those interested in the next step in scaling ML performance, this will likely be it. GPUDirect will enable direct SSD to GPU mem access, bypassing the CPU entirely; while RAPIDS will provide seamless data loaders to consume that data into data frames, ML algos, etc.
This week NVIDIA unveiled GPUDirect Storage - a move to continue to expand its reach in #datascience applications like @RAPIDSAI. Read more in this exclusive interview with our GM of Data Science, @datametrician. nvda.ws/2MLKPdU
Gemma 2 is out!
As with our first model, we're super focused on creating models at useful, practical sizes, so that they can be easily deployable... all the while being amazing in quality.
We upgraded our 9B so that it's truly awesome and best in class across many benchmarks.
Today we released Gemma, our latest open models. 2B and 7B, best in class. You'll find them anywhere, try them on HF huggingface.co/chat?model=goo….
Want to congratulate our amazing team that made it possible, building on everything Gemini has built. Special congrats to
Gemini Deep Research mode is one of the features we released, that's super useful day to day. Here I just asked for a summary of how our new Gemini Multimodal Live API got received... it crawled the web, built up a great report, with references, and I can just share it via
Excellent post from @karpathy — I've been making a similar point in a recent talk. Deep Learning is essentially redefining how we write software: source code is being replaced by data, compilers by deep learning, and executables by trained predictors. x.com/karpathy/statu…
Thread (personal view). I watched the Social Dilemma on Netflix. Highly recommend. It brought back many good and bad memories... and mostly made me very uneasy. Most folks you see on this video have profited tremendously from the "attention selling" industry. Back 10y ago, it was
Woah, huge news again from Chatbot Arena🔥
@GoogleDeepMind’s just released Gemini (Exp 1121) is back stronger (+20 points), tied #1🏅Overall with the latest GPT-4o-1120 in Arena!
Ranking gains since Gemini-Exp-1114:
- Overall #3 → #1
- Overall (StyleCtrl): #5 -> #2
- Hard