RUN THIS LLM
Search local LLM hardware requirements
EmbeddingGemma 2
Google · 740M · Embedding
Google DeepMind's multimodal on-device embedding model built on Gemma 4 — text, code, images, video, and audio in one 768-dim space, 100+ languages, 8K context, Matryoshka truncation down to 128 dims. 740M total: a 270M text backbone plus optionally loaded 170M vision and 300M audio encoders.
VRAM Requirements
| Quantization | VRAM |
|---|---|
| FP16 (full quality) | 1.5 GB |
Specifications
- Parameters: 740M
- Category: Embedding
- System RAM: 2 GB minimum
- HuggingFace: google/embeddinggemma-2
Benchmarks
- MTEB: 61 — Embedding retrieval quality (avg)
Loading interactive analysis...