Google open-sources EmbeddingGemma 2, a 740M multimodal embedding model that runs in-browser
DynamicWebPaige · x · 2026-10-07
Google released EmbeddingGemma 2, its first open, natively multimodal embedding model. A 740M-parameter model handles text, code, image, video and audio, claims to outperform specialist models twice its size, and pairs with Gemma 4 for offline privacy-first RAG. Weights are live on Hugging Face.
Developers report it runs on-device in the browser via WebGPU with no server or API key, at 20–70ms per query, keeping all data local.
More from Models
- Claude was given robot skin — Opus was curious but anxious about hooking up — repligate · 2026-10-07
- Gemini 2.5 Pro retiring October 20, 2026, users say goodbye — hargup13 · 2026-10-07
- OpenAI shipped Decisions API just two weeks after first prototype — stevenheidel · 2026-10-07
- User marvels as ChatGPT shows humanlike academic skills — dioscuri · 2026-10-07
- Why Hybrid Models May Scale Better Downstream: The Inductive-Bias Argument — kalomaze · 2026-10-07
- Four LLMs paired in a co-op game: strong+strong wins, humans still beat AI — abrdeveloper · 2026-10-07