Armand Joulin cohere Thanks Armand Joulin -- we will make a correction. We group all private testing by provider. So while overall number of variants is correct, in this case there is very different testing patterns per model family under a provider. We will clarify gemma only had one private test.

Qwen3 validating on policy distillation at scale for thinking!

We’ve developed Gemini Diffusion: our state-of-the-art text diffusion model. Instead of predicting text directly, it learns to generate outputs by refining noise, step-by-step. This helps it excel at coding and math, where it can iterate over solutions quickly. #GoogleIO

1000+ words per second! ⚡ We just unleashed Gemini Diffusion at #GoogleIO! 🚀 Awesome being part of the team that took this from a small research project all the way to I/O Google DeepMind 🪐

It's been an incredible experience being part of the team that took this from a small research project all the way to I/O 🪐 Super proud of the team! Google DeepMind

🚀Meet Gemini Diffusion, our first diffusion-based and super fast language model, just announced at Google I/O!🚀 Very excited to be able to share what I've been working on for the past little while with our amazing small team Google DeepMind.

Releasing Gemma 3n, our new open-weight model processing audio, images and text (with improved multilingual capabilities), optimized for on-device usage with MatFormer architecture (enabling adaptive compute) and reaching 1283 on Chatbot Arena. Read more: developers.googleblog.com/en/introducing….

🔥 Gemini Diffusion is blazing fast 🔥 Honored to have been part of this amazing team!

Gemini Diffusion has been announced at #GoogleIO! 🚀 Diffusion for text allows for self correction and incredibly fast inference by generating tokens in parallel over a long horizon. Super proud to have played a small part in making this happen over the last two years.

Excited to share what I've been up to: Gemini Diffusion is FAST! I'm convinced this will revolutionise iterative workflows: refine, get instant feedback, repeat! So proud of what our small team achieved here🪐

Gemini Diffusion is out! Very excited to have worked on the post-training of such a state-of-the-art text diffusion model. Incredible performance at lightspeed⚡️ Congrats to everyone involved!!

Pocket powerhouse admist I/O awesomeness! Gemma 3n E4B & E2B are insane models, optimized for on-device while rivaling frontier models. It's a 🪆Matryoshka Transformer (MatFormer)🪆: Natively elastic b/w 4B & 2B pareto-optimally! ⭐️: free models with ZERO training cost! 🧵👇

Very excited to share what I have been working on. Having been part of the Gemini Diffusion team since day one, it is amazing to see our model demoed at Google I/O :) sign up below to try it out!

Excited to share what my team has been working on lately - Gemini diffusion! We bring diffusion to language modeling, yielding more power and blazing speeds! 🚀🚀🚀 Gemini diffusion is especially strong at coding. In this example the model generates at 2000 tokens/sec,

Such a privilege to work on Gemini Diffusion with an amazing team! From a small research project to launching at I/O - we've got unstoppable aura 🚀 Welcome to the era of live vibe coding ⚡️

Really proud that two new models have been presented at I/O which we have post-trained: - Gemini Diffusion: with >1k tokens per second a completely new LLM experience deepmind.google/models/gemini-… - Gemma 3n: pushing the boundary of what is possible on mobile developers.googleblog.com/en/introducing…

Gemma 3n is out! 🚀🚀🚀 The frontier models from a year ago can now run locally on a phone! Lots of innovations (e.g. matformers, mix’n’match, per layer embeddings) to make this model mobile first. And we finally have audio/video as an input for Gemma models! 1/2

Gemma 3n 🚀🚀🚀

Gemma 3n E4B has the same number of total parameters (8B) as the original Gemma 7B (8B lol). The progress of Gemma over the past 16 months is insane

Check out our latest release: It's exciting that on-device models can be that strong!