Google releases EmbeddingGemma 2: a 740M-parameter natively multimodal open embedding model
GoogleDeepMind · x · 2026-10-07
Google DeepMind released EmbeddingGemma 2, its first natively multimodal open model for on-device embeddings, under Apache 2.0 with weights on Hugging Face and Kaggle.
- Size & performance: only 740M parameters yet competitive across benchmarks, outperforming some specialist models over twice its size.
- Unified modality space: extends beyond text to code, images, audio, and video.
- Use cases: multimodal search in apps (e.g., finding video moments from a voice memo) or pairing with Gemma 4 for private on-device RAG.
More from Models
- Grok's quirk: it says 'No.' then argues your point better than you did — gandamu_ml · 2026-10-07
- Phil Schmid demos embedding anything with Google Gemma directly in the browser — _philschmid · 2026-10-07
- Anthropic offers Pro users $100 and Max users $250 bonus credit for Claude Code cloud sessions — ClaudeDevs · 2026-10-07
- vLLM adds day-0 support for Google's multimodal EmbeddingGemma 2 — AccBalanced · 2026-10-07
- RedNote Has a Serious, Low-Key AI Lab, Argues Analyst Weighing Dots3 Results — teortaxesTex · 2026-10-07
- Mistral Large 4 reportedly beats Qwen 3.8 Max and Kimi K3 on Terminal-Bench — Evermoving- · 2026-10-07