Gemma 3 100% Local using @ollama and @chainlit_io Create an Application DEMO: 👇 🚀 27B, 12B, 4B, 1B 🌐 140+ languages supported 📊 128K context window 🖼️ Multimodal capabilities 🏆 Outperforms Llama 405B, DeepSeek V3 💻 Runs on single GPU 🤖 Enables function calling 🔍 Vision understanding capability 🧠 14T training tokens 🔄 5:1 local-global ratio 📱 Local processing possible 🎯 67.5 MMLU-Pro score 🏅 133 Elo score 📦 Int4 quantisation support @GoogleAI @GoogleDeepMind @sundarpichai
See Tweet