EmbeddingGemma Models by Google DeepMind
Google DeepMind releases · 52 in the last 12 months · this family litChangelog →
1 model2026Up to 8k ctx
Last refreshed 2026-10-07. Next refresh: weekly.
Details
ResearcherGoogle DeepMind
LicenseApache 2.0OSI-approved
Commercial useCommercial use: permitted
Models1
Released2026
Max context8k
Capabilities
VisionAll models
MultimodalAll models
About
EmbeddingGemma is Google DeepMind's family of open-weight embedding models built on Gemma technology for on-device retrieval, search, and RAG. EmbeddingGemma 2 (October 2026) is natively multimodal, mapping text, code, images, video, and audio into one 768-dimensional space, with 740M total parameters and Matryoshka truncation to 512, 256, or 128 dimensions. Weights are released under Apache 2.0 on Hugging Face and Kaggle.
Decision facts
- Best fit
- embeddingvision and multimodal workcoding
- Capability starting point
- EmbeddingGemma 2 with 8k context and multimodal inputs
- Lowest tracked input
- Not tracked
- Closest related family
- DiffusionGemma
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
1 in view
EmbeddingGemma 2Current
Use when the workload needs embedding, 8k context, and 740M parameters.
2026-10embedding8k context740M parameters
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| EmbeddingGemma 2 | Use when the workload needs embedding, 8k context, and 740M parameters. | 2026-10 | embedding8k context740M parameters | Current |
Release Timeline
1 release group2026-10
1 current
EmbeddingGemma 2
Currentembedding8k context740M parameters
Specifications(1 models)
| Model | Released | Context | Parameters | Vision | Multimodal |
|---|---|---|---|---|---|
| EmbeddingGemma 2 | 2026-10 | 8k | 740M | Yes | Yes |