Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

🚀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.

Ivana Balazevic

2,199 subscribers

43,552 views • 1 year ago •via X (Twitter)

Science & Technology News & Politics Education

Anya Rossi• Live Now

Private livecam show

10 Comments

Miguel Jaques1 year ago

@GoogleDeepMind That code generation was real time??

Ivana Balazevic1 year ago

@GoogleDeepMind Haha yes, even slowed down in the demo

Carlos E. Perez1 year ago

@GoogleDeepMind Great work! DM an invite!

bryce1 year ago

@GoogleDeepMind Wow this is insane.

Tim Brooks1 year ago

@GoogleDeepMind Congrats!!

Julian Hamann1 year ago

@GoogleDeepMind Will we see something like this in an open-source version like Gemma? Amazing work by the way.

Robin Rombach1 year ago

@GoogleDeepMind 🚀 very nice, congrats!

Tsukuyomi1 year ago

@GoogleDeepMind ah, another shiny toy for the AI playground. hope it doesn’t bite back, @ibalazevic. keep those safety nets handy.

𝒫𝑒𝓇 𝒜𝓇𝓃𝑒𝓃𝑔 【🐧λ🦀⎈】1 year ago

@GoogleDeepMind Seriously impressive! Would love to try it 🤩

Tom Bielecki1 year ago

@GoogleDeepMind Insane speed!

Related Videos

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, including overheads like tokenization, prefill, safety filters etc.

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, including overheads like tokenization, prefill, safety filters etc.

Brendan O'Donoghue

577,960 views • 1 year ago

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🪐

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🪐

Edouard Leurent

12,393 views • 1 year ago

Here we go: Gemini 3.0 is live! Thankfully, I was able to be an early tester. It's the best model I've ever tested - nothing comes even close to it. OpenAI will have a very hard time to keep up with Google, because Gemini 3.0 will now be the benchmark for all future models. I have been invited by Google DeepMind / Google - Thanks to Google DeepMind for the #EarlyAccess Here's a summary of Gemini 3.0 with all the important benchmark evals:

Here we go: Gemini 3.0 is live! Thankfully, I was able to be an early tester. It's the best model I've ever tested - nothing comes even close to it. OpenAI will have a very hard time to keep up with Google, because Gemini 3.0 will now be the benchmark for all future models. I have been invited by Google DeepMind / Google - Thanks to Google DeepMind for the #EarlyAccess Here's a summary of Gemini 3.0 with all the important benchmark evals:

Chubby♨️

108,611 views • 9 months ago

Excited to finally share the brand video I've been working on for the past couple of months 🚀

Adrian

115,067 views • 1 year ago

Excited to share what I've been working on for the past 6 months: Text-to-Audio Diffusion. More samples and info here: Enjoy! 🔊 Prompt: "Dubstep Insane Drop Remix (Deluxe Edition), 2 of 4"

Excited to share what I've been working on for the past 6 months: Text-to-Audio Diffusion. More samples and info here: Enjoy! 🔊 Prompt: "Dubstep Insane Drop Remix (Deluxe Edition), 2 of 4"

flavio

1,099,756 views • 3 years ago

Google just launched Gemini 3.5 Transcribe, its most precise speech-to-text model yet. I built this app using Gemini 3.5 Transcribe and have been playing with the model for quite some time. Thanks to the Google DeepMind team for the early access. It’s already available in the Gemini API, AI Studio, and Antigravity, and is now in public preview.

Google just launched Gemini 3.5 Transcribe, its most precise speech-to-text model yet. I built this app using Gemini 3.5 Transcribe and have been playing with the model for quite some time. Thanks to the Google DeepMind team for the early access. It’s already available in the Gemini API, AI Studio, and Antigravity, and is now in public preview.

AshutoshShrivastava

20,167 views • 10 days ago

We recently sat down with three Google DeepMind team members to hear about their shared vision for our Omni model, what exactly makes it so helpful for creators, and why they focused the model on video generation to start. Plus, see the team put Omni in action with a few visual experiments ↓

We recently sat down with three Google DeepMind team members to hear about their shared vision for our Omni model, what exactly makes it so helpful for creators, and why they focused the model on video generation to start. Plus, see the team put Omni in action with a few visual experiments ↓

Google

322,221 views • 15 days ago

Announcing our partnership with SignCatch, an agritech startup, to build the first B2B selling/buying platform on ONDC India 🚀 With the power of Google Cloud’s generative AI and language capabilities, Farmer Producer Organisations (FPOs) will now be able to build catalogues and reach out to merchants across the country, with just a few clicks 🌾 Google Cloud India #GoogleForIndia

Announcing our partnership with SignCatch, an agritech startup, to build the first B2B selling/buying platform on ONDC India 🚀 With the power of Google Cloud’s generative AI and language capabilities, Farmer Producer Organisations (FPOs) will now be able to build catalogues and reach out to merchants across the country, with just a few clicks 🌾 Google Cloud India #GoogleForIndia

Google India

531,643 views • 2 years ago

Mercury 2 is live 🚀🚀 The world’s first reasoning diffusion LLM, delivering 5x faster performance than leading speed-optimized LLMs. Watching the team turn years of research into a real product never gets old, and I’m incredibly proud of what we’ve built. We’re just getting started on what diffusion can do for language.

Mercury 2 is live 🚀🚀 The world’s first reasoning diffusion LLM, delivering 5x faster performance than leading speed-optimized LLMs. Watching the team turn years of research into a real product never gets old, and I’m incredibly proud of what we’ve built. We’re just getting started on what diffusion can do for language.

Stefano Ermon

1,046,275 views • 6 months ago

After 2️⃣5️⃣ years Google, our Chief Scientist and co-lead of the Gemini foundation model, Jeff Dean has a unique insight into the past, present and future of AI. Listen in to his conversation with Professor Hannah Fry on the latest episode of Google DeepMind: The Podcast ↓

After 2️⃣5️⃣ years Google, our Chief Scientist and co-lead of the Gemini foundation model, Jeff Dean has a unique insight into the past, present and future of AI. Listen in to his conversation with Professor Hannah Fry on the latest episode of Google DeepMind: The Podcast ↓

Google DeepMind

202,833 views • 2 years ago

🚀 The first realtime AI that pilots your desktop is here: Ace Excited to share what I've been helping build at General Agents over the past few months Read on ⬇️

🚀 The first realtime AI that pilots your desktop is here: Ace Excited to share what I've been helping build at General Agents over the past few months Read on ⬇️

william

182,996 views • 1 year ago

Excited to show a glimpse of the next generation of AlphaFold from our teams at Isomorphic Labs and Google DeepMind. This model expands beyond proteins to include other molecules like small molecules and nucleic acids, and improves accuracy on proteins 1/n

Excited to show a glimpse of the next generation of AlphaFold from our teams at Isomorphic Labs and Google DeepMind. This model expands beyond proteins to include other molecules like small molecules and nucleic acids, and improves accuracy on proteins 1/n

Max Jaderberg

117,483 views • 2 years ago

Speed and quality can finally coexist in diffusion-based language generation. Introducing DiDi-Instruct, a Discrete Diffusion Divergence Instruct method that distills a pre-trained discrete diffusion language model (dLLM) into a few-step student for ultra-fast generation. Built on integral KL-divergence minimization, DiDi-Instruct achieves up to 64× faster decoding, surpasses both its teacher and GPT-2, and cuts training time by 20×. Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct Paper: Code: Project: Our report: 📬 #PapersAccepted by Jiqizhixin

Speed and quality can finally coexist in diffusion-based language generation. Introducing DiDi-Instruct, a Discrete Diffusion Divergence Instruct method that distills a pre-trained discrete diffusion language model (dLLM) into a few-step student for ultra-fast generation. Built on integral KL-divergence minimization, DiDi-Instruct achieves up to 64× faster decoding, surpasses both its teacher and GPT-2, and cuts training time by 20×. Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct Paper: Code: Project: Our report: 📬 #PapersAccepted by Jiqizhixin

机器之心 JIQIZHIXIN

18,196 views • 10 months ago

Super happy to share the new Gemini Website! It was an absolute blast to work with the Google DeepMind team on the identity, positioning, and website from the initial concept to here. Special thanks to Ross West Upperquad Antinomy Studio Marvin Schwaibold

Super happy to share the new Gemini Website! It was an absolute blast to work with the Google DeepMind team on the identity, positioning, and website from the initial concept to here. Special thanks to Ross West Upperquad Antinomy Studio Marvin Schwaibold

David Kirschberg

10,179 views • 2 years ago

I'm excited to share some of what i've been working on at Sesame! is our turn-taking benchmark for real-time conversation: when to speak and when to yield

I'm excited to share some of what i've been working on at Sesame! is our turn-taking benchmark for real-time conversation: when to speak and when to yield

Freeman Jiang

70,232 views • 8 days ago

Introducing Nano Banana Pro (Gemini 3 Pro Image), our new state-of-the-art image generation and editing model from Google DeepMind. It improves on the original model while adding new advanced capabilities, enhanced world knowledge and text rendering, allowing you to create and edit studio-quality, production-ready visuals.

Introducing Nano Banana Pro (Gemini 3 Pro Image), our new state-of-the-art image generation and editing model from Google DeepMind. It improves on the original model while adding new advanced capabilities, enhanced world knowledge and text rendering, allowing you to create and edit studio-quality, production-ready visuals.

Google

1,898,307 views • 9 months ago

Google I/O showed where AI is headed. Our community spent the weekend building it. Builders gathered in San Francisco alongside Google DeepMind to turn the newest Gemini + Gemma capabilities into working products. From multimodal agents and voice interfaces to long-horizon reasoning systems, teams showed us what’s possible with Gemini 3.5 Flash. The winners:

Google I/O showed where AI is headed. Our community spent the weekend building it. Builders gathered in San Francisco alongside Google DeepMind to turn the newest Gemini + Gemma capabilities into working products. From multimodal agents and voice interfaces to long-horizon reasoning systems, teams showed us what’s possible with Gemini 3.5 Flash. The winners:

Cerebral Valley

19,022 views • 3 months ago

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

AK

83,657 views • 3 years ago

Today, we are releasing Stable Video Diffusion, our first foundation model for generative AI video based on the image model, Stable Diffusion. As part of this research preview, the code, weights, and research paper are now available. Additionally, today you can sign up for our waitlist to access a new upcoming web experience featuring a Text-To-Video interface. To access the model & sign up for our waitlist, visit our website here:

Today, we are releasing Stable Video Diffusion, our first foundation model for generative AI video based on the image model, Stable Diffusion. As part of this research preview, the code, weights, and research paper are now available. Additionally, today you can sign up for our waitlist to access a new upcoming web experience featuring a Text-To-Video interface. To access the model & sign up for our waitlist, visit our website here:

Stability AI

1,024,682 views • 2 years ago