---
language:
- en
- multilingual
license: gemma
base_model: google/embeddinggemma-300m
tags:
- openvino
- feature-extraction
- sentence-similarity
- sentence-transformers
- text-embeddings-inference
library_name: sentence-transformers
pipeline_tag: feature-extraction
---

# EmbeddingGemma-300M OpenVINO FP32

OpenVINO IR (FP32) export of [google/embeddinggemma-300m](https://huggingface.co/google/embeddinggemma-300m),
packaged for [IREmbeddingServer](https://github.com/) inference (`MODEL_ID=embeddinggemma-300m-ov`).

## Layout

| Path | Description |
|------|-------------|
| `openvino_model.xml` / `openvino_model.bin` | Gemma3 backbone IR (~1.2GB FP32) |
| `2_Dense/` / `3_Dense/` | Sentence Transformers Dense layers (post-OV matmul) |
| `tokenizer.json` | Tokenizer |
| `modules.json` | ST module metadata |

## IREmbeddingServer

```bash
export EMBED_MODEL_DIR="./embeddinggemma-300m-ov"
export MODEL_ID=embeddinggemma-300m-ov
export EMBED_DEVICE=cpu   # or igpu / npu
```

## License

Derived from Google's EmbeddingGemma (Gemma license). See [google/embeddinggemma-300m](https://huggingface.co/google/embeddinggemma-300m).