EmbeddingGemma 2: Specs, License & How to Run
Google DeepMind open multimodal embedding model (~0.74B, Apache-2.0).
TL;DR. EmbeddingGemma 2 is a embedding model from Google DeepMind. Parameters: ~0.74B.License: Apache-2.0. It is listed under Closed embedding APIs.
| Developer | Google DeepMind |
|---|---|
| Type | embedding |
| Architecture | dense |
| Parameters | ~0.74B |
| Modalities | text, code, image, video, audio |
| License | Apache-2.0 |
| Hardware | Not yet verified — On-device VRAM/RAM figures were not in the seed extract. (not independently tested) |
| Hugging Face likes | 1,889 (as of 2026-10-13) |
| Model card | https://huggingface.co/google/embeddinggemma-2 |
| First seen | 2026-09-14 |
EmbeddingGemma 2 is Google DeepMind’s open multimodal embedding model. It is grouped under open alternatives to closed embedding APIs.
Specs (2026-10-13)
- Card: https://huggingface.co/google/embeddinggemma-2
- ~0.74B parameters
- Modalities: text, code, image, video, audio
- License: Apache-2.0
- 1,889 likes
- Listed 2026-09-14
On-device
A verified consumer of the model is DigUp, a Mac local semantic file search application (https://github.com/ARahim3/DigUp, MIT, Swift, 1,712★, created 2026-10-08). For a generic run outline, see Run EmbeddingGemma 2 on device (docs-based, not yet tested by forkalt).
Compare
How to run
- How to Run EmbeddingGemma 2 On-Device (based on official docs, not yet tested)
Deploy options
- local: Hugging Face model card
Comparisons
FAQ
What modalities does EmbeddingGemma 2 support?
Text, code, image, video, and audio, per the Hugging Face card (2026-10-13).
How large is it?
About 0.74B parameters, per the same card.
Is there an app that already uses it?
DigUp (ARahim3/DigUp) is a Mac on-device semantic file search app built on EmbeddingGemma 2 (1,712 GitHub stars, MIT, Swift, created 2026-10-08). That relationship is verified; we did not give DigUp its own seed URL.
How does it compare to BGE-M3?
See EmbeddingGemma 2 vs BGE-M3. We do not quote MTEB numbers we did not source.
Sources & verification
License notes are not legal advice. Read the original LICENSE. Trademark & disclaimer.