Google open-sources EmbeddingGemma 2: on-device multimodal search across photos, audio, docs
bigaiguy · x · 2026-11-02
Google open-sourced EmbeddingGemma 2, letting phones search photos, videos, voice notes and documents with one query — fully offline, no cloud upload or pre-captioning/transcription. Key specs: maps text, code, image, video and audio into one shared embedding space; 740M params; 567MB RAM for full multimodal on a Pixel (191MB text-only); 8K token context (4x the first version); 100+ languages; embeddings shrinkable to 512/256/128 dims; INT4/INT8 builds for edge devices.
More from Models
- Nanbeige4.3-33B lands on Hugging Face, pitched as a Qwen rival — amaarora · 2026-11-02
- A4 print-ready PDF stress test: only Claude, Gemini and Mistral pass out of 10 AI models — gerardsans · 2026-11-02
- Discounted API relay stations lack prompt caching, making long sessions pricier — uneatenbreakfast · 2026-11-02
- Anthropic's server-side rule reignites the case for owning open models over renting AI — alex_verem · 2026-11-02
- Epoch AI: open-weight GLM 5.3 exploited the patch gap to hack 6 Firefox releases at ~$1,000 each — rohanpaul_ai · 2026-11-02
- SemiAnalysis: safety classifiers sandbag scores of Opus 5.5 and GPT-6 Astra — morqon · 2026-11-02