# Google publishes the EmbeddingGemma 2 technical report

> Original title: The EmbeddingGemma 2 tech report is officially live!

- Source: Google Gemma
- Published: 2026-11-05T16:36:59.000Z
- Bad Influence: https://badinfluence.news/items/alrit29qptylkha6wxlewlyin
- Original: https://x.com/googlegemma/status/2118382075564929178

## Summary

Google says it has published the technical report for EmbeddingGemma 2, a 740M-parameter open multimodal embedding model. The report covers its architecture, training recipe, and ablations. Google says the model represents text, code, images, audio, and video in one space and outperforms specialist and general models up to 12x larger. It also lists on-device options such as modular encoders, Matryoshka dimensions, and int4 quantization down to 567MB.

## Why it matters

The report gives architecture, training recipe, and ablations for an open multimodal embedding model, useful for judging how unified embeddings trade accuracy against on-device size.

## Post

The EmbeddingGemma 2 tech report is officially live!

We are sharing the full architecture, training recipe, and ablations behind our 740M open multimodal embedding model.

Inside the paper:
🌐 Unified representations for text, code, images, audio, and video
📈 How it outperforms specialist and general models up to 12x larger
⚡ On-device efficiency: modular encoders, Matryoshka dims, and int4 quantization (down to 567MB)

Read the full report: https://arxiv.org/abs/2610.37364

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWYBPdya0AEp6uj.png&mode=thumb&exp=1794096000&sig=ea23980acbbdbfad)