EmbeddingGemma — local AI model by Google DeepMind

Written by Jakub Rusinowski · Last updated

Google's on-device embedding models for search, RAG and clustering. EmbeddingGemma 2 (October 2026) puts text, code, images, video and audio into one 768-dimension vector space that can be shortened to 512, 256 or 128 dimensions. The full model has 740M parameters, but the text-only setup loads 270M, so it runs comfortably on laptops and phones. Apache 2.0. An embedding model writes no text: it is the retrieval half of a RAG setup, next to a chat model.

Variants

The smallest EmbeddingGemma variant needs about 2 GB of VRAM at BF16 — quantized weights plus framework overhead, before any KV cache.

ModelVRAM
EmbeddingGemma 2 →
740M (270M text-only)
~2.3 GB

Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.

How to run EmbeddingGemma locally

Install Ollama, then pull the tag.

Ollama carries it as embeddinggemma-2 (270m text-only, 440m with images, 570m with audio, 740m with everything). Pull it with `ollama pull embeddinggemma-2:270m` and call it through the embeddings API (ollama.embed); an embedding model is not chatted with, so there is no `ollama run` command.

Pick a size above for its own VRAM figure, speed estimate and install command.

Licence

Apache-2.0Commercial use permitted

Commercial use permitted. No usage restrictions beyond attribution.

Applies to: EmbeddingGemma 2

Recommended GPU

The cheapest catalogued GPU that runs EmbeddingGemma locally (min 2 GB VRAM) is the Intel Arc B570 (10 GB).

Affiliate disclosure: Some links on this page are affiliate links — if you buy through them, LLM Configurator may earn a commission at no extra cost to you. As an Amazon Associate, LLM Configurator earns from qualifying purchases.
Intel Arc B570 10GB
10 GB VRAM · 150 W board power
2026 prices are volatile — check the current listing.

EmbeddingGemma — frequently asked questions

How much VRAM does EmbeddingGemma need?

EmbeddingGemma needs about 2 GB VRAM at BF16 quantization for its smallest variant. Variants: EmbeddingGemma 2 (2 GB, BF16). On Apple Silicon, unified memory counts toward this requirement.

Can I run EmbeddingGemma on an RTX 4090 (24 GB)?

Yes — EmbeddingGemma runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at BF16.

What quantization should I use for EmbeddingGemma?

EmbeddingGemma is published only pre-quantised, so there is no Q4_K_M or Q8_0 build to choose between: each variant ships in the format listed beside it above (BF16). Use the vendor's runtime for that format, as the variant pages describe.

How do I run EmbeddingGemma with Ollama?

EmbeddingGemma is an embedding model: pull it with `ollama pull embeddinggemma-2:270m` and call it through Ollama's embeddings API. It turns text into vectors for search and RAG and is not chatted with, so there is no `ollama run` command.