EmbeddingGemma 2: Google DeepMind's Compact Multimodal Embedding Model

LinkGoogle DeepMindTool

Google DeepMind has released EmbeddingGemma 2, an open multimodal embedding model with 740M parameters that maps text, code, images, video, and audio into a unified 768-dimensional vector space.

Key points

  1. Total of 740M parameters, combining a 270M text backbone with modular vision (170M) and audio (300M) encoders.
  2. Built on Gemma 4 architecture, offering support for 100+ languages and an 8,192-token context window.
  3. Features Matryoshka Representation Learning (MRL) supporting dimensions of 128, 256, 512, and 768 for up to 6x storage reduction.
  4. Selective encoder loading allows omitting unused modalities to shrink memory footprint down to 270M for text-only workloads.
  5. Employs lightweight task instruction prefixes to steer embeddings for asymmetric and symmetric tasks.
  6. Must be run using bfloat16 or float32; float16 causes activations to exceed dynamic range, leading to NaNs or silent degradation.

Why it matters

It enables unified on-device semantic search and multimodal RAG directly on consumer devices without the overhead of coordinating separate embedding models across media formats.

Analysis

The modular architecture and MRL support give teams high operational flexibility. Hardware footprints can be tailored by dropping unnecessary modality encoders, and vector storage costs can be trimmed significantly by truncating dimensions down to 256d with minimal quality degradation.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 7 October 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:Mia

    Google Releases EmbeddingGemma 2: Unified Multimodal Vector Space

    Google has launched EmbeddingGemma 2, an open-weights model mapping text, code, images, video, and audio into a single shared vector space. The 1.49 GB model is released under the Apache 2.0 licence.

    684bookmarks37kviews

    AI & LLMs· Announcement

  • Article:Nikita Pavlichenko

    JetBrains Introduces Mellum2: A 12B Mixture-of-Experts Model for Code and Text

    JetBrains has released Mellum2, a 12B-parameter Mixture-of-Experts model trained from scratch on natural language and code. Activating only 2.5B parameters per token and published under the Apache 2.0 license, it is designed for low-latency, high-throughput developer and AI workloads.

    AI & LLMs· Announcement

  • Repository:drumih/turbo-fieldfare

    TurboFieldfare: Gemma 4 26B-A4B Inference in 2 GB RAM on Apple Silicon

    TurboFieldfare is a dedicated Swift and Metal runtime designed to run Google's Gemma 4 26B-A4B on Apple Silicon Macs within an approximately 2 GB RAM budget. It avoids loading the full 14.3 GB model into memory by keeping only the 1.35 GB core and KV cache in RAM while streaming the required mixture-of-experts weights from SSD on demand.

    6.6kstarsSwift

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.