Loading…
17-18 September | Amsterdam, Netherlands
View More Details & Registration

IMPORTANT NOTE: Timing of sessions and room locations are subject to change.
Thursday September 17, 2026 09:40 - 09:50 CEST
In this talk, I will use a LOT of visuals to go through some of the (fundamental) architectural innovations that were used in the Gemma 4 models that make them ideal for local usage:

* Per-Layer Embeddings (PLE) - Adding parameters without adding them
* Multi-Token Prediction (MTP) drafters - Use smaller models to generate candidate tokens
* Mobile QAT - Making E2B/E4 even more efficient
* Gemma 4 12B - Doing away with the encoders
* DiffusionGemma - A diffusion-tuned Gemma 4 26B
Speakers
avatar for Maarten Grootendorst

Maarten Grootendorst

Senior Developer Relations Engineer, Google DeepMind
Maarten Grootendorst is a Senior Developer Relations Engineer at Google DeepMind. He holds master’s degrees in organizational psychology, clinical psychology, and data science, which he leverages to communicate complex machine learning concepts to a wide audience. With his popular... Read More →
Thursday September 17, 2026 09:40 - 09:50 CEST
Auditorium (Ground Level)

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Share Modal

Share this link via

Or copy link