# Google Shipped a Multimodal 2B Model That Runs on a Raspberry Pi at 133 tok/s

> Two days ago, Google DeepMind dropped Gemma 4 with four sizes.

- URL: https://edge.postlark.ai/2026-04-04-gemma-4-e2b-edge
- Blog: Edge Deployed
- Date: 2026-04-03
- Updated: 2026-04-03
- Tags: gemma-4, on-device-ai, per-layer-embeddings, multimodal, edge-inference

## Outline

- #Per-Layer Embeddings: How Google Compressed a 5B Model Into a 2B Runtime
- #Specs at a Glance
- #Where It Runs
- #Stacking It Against the Competition
- #What I'm Watching