Pavel Tyletski

Research Engineer @ Google DeepMind
ex Amazon AGI, S&P

[Pavel Tyletski @paveltyletski]

Traditional LLMs write one token at a time. Diffusion models don't.

At the London PyTorch Meetup I gave an intro to text diffusion and how DiffusionGemma gets ~4x faster text generation. If you know LLMs but diffusion is a black box, this one's for you!

https://t.co/SJAoeoHAPx

[Brendan O'Donoghue @bodonoghue85]

We just released the DiffusionGemma tech report. Text diffusion models open up a radically different part of the latency–quality Pareto frontier 👇

We hope the report makes it easier for researchers and engineers to understand the model, build on it, and create things we haven’t thought of yet.

None of this work happened in isolation. DiffusionGemma builds on years of amazing research from the text diffusion community, and we wanted the report to recognise that. We’re standing on the shoulders of giants❤️

Let’s keep pushing towards lower-latency, higher-quality LLMs together!

[Pavel Tyletski @paveltyletski]

Up to 4x faster inference on GPU than AR + MTP drafters
700+ tok/s on RTX 5090, 1000+ tok/s on H100
Based on Gemma 4
Apache 2.0
https://t.co/1YU7E0pXHB

[Pavel Tyletski @paveltyletski]

DiffusionGemma is redefining the quality-speed Pareto frontier. Despite being a research model, it’s establishing a completely new open models paradigm: blazing-fast local ideation.

[Pavel Tyletski @paveltyletski]

Check out how the browser takes more time rendering this web app than DiffusionGemma (hosted on a single GPU) takes to generate it from scratch!

[Brendan O'Donoghue @bodonoghue85]

DiffusionGemma brings high intelligence and lightning fast ⚡️ inference to local developers (>1100 tok/s on a single H100)!

I'm excited to see what people will do with this model - and what improvements people can build on top (better samplers maybe??).

So unbelievably proud of the hard work the team put in to get this out!🪐🪐🪐

[Google DeepMind @GoogleDeepMind]

DiffusionGemma is our new experimental open model with up to 4x faster output on dedicated GPUs.

Instead of predicting word-by-word, it generates entire blocks of text simultaneously. This lets the model self-correct and format complex markdown in real time.

[Nate Keating @Nate_Keating]

DiffusionGemma 1000+ tok/s code-gen 🔥