Developers building private, on-device search across text, images and audio get it free under Apache 2.0.
Kenning's take
What it does
EmbeddingGemma 2 is a 740 million parameter open embedding model from Google that maps text, code, images, audio and video into one shared space. Developers use it for search and retrieval that runs on phones and laptops without sending data to a server. It adds an 8K context, four times the first version, and is released under Apache 2.0 on Hugging Face and Kaggle.
- Text, code, image, audio and video embeddings in one space
- 740M parameters, or 270M for text only
- 8K token context
- Vectors can be cut from 768 to 128 dimensions
- Runs with Transformers, llama.cpp, Ollama, MLX and vLLM
Who it suits. Developers building on-device search, retrieval and RAG features.
Key facts
- Category
- AI models
- Made by
- Google DeepMind
- Released
- 6 October 2026
- How to get it
- open weights on Hugging Face and Kaggle
- Works on
- open weights
- In the UK
- Yes
- Free plan
- open weights, Apache 2.0
FAQ
When was EmbeddingGemma 2 released?
Google announced it on 6 October 2026.
What is new in EmbeddingGemma 2?
It handles images, audio, video and code as well as text, and has an 8K token context.
Is EmbeddingGemma 2 free?
Yes. It is released under the Apache 2.0 licence on Hugging Face and Kaggle.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.