The Allen Institute for AI (AI2) has launched a new open-source iOS app that enables local inference of its updated OLMoE AI models directly on mobile devices. This development allows users to run AI models securely and privately without relying on cloud-based servers. The app supports the latest OLMoE-1B-7B-0125 model, which incorporates advancements like the Tulu 3 recipe and RLVR, improving performance metrics such as gsm8k by nearly 10 points. The app demonstrates the potential of on-device AI, providing faster processing and enhanced privacy, with speeds of approximately 20 tokens per second on an iPhone 16. AI2’s innovation aligns with the growing trend of making AI more accessible and practical for everyday use.
Chat with the world’s smartest AI completely privately on your Apple Mac or Windows computer. You download the app once, and chat forever with no limits on usage, no subscriptions, and 100% privacy. https://t.co/bWBZsennJG
Now you can chat with the world’s smartest AI privately on your Apple Silicon, Intel Mac, or Windows machine. You download the app once, and chat forever with no limits on usage and no hidden fees. Your conversations stay local, secure, and private – we can’t access them and… https://t.co/pnrGsjnUsV
Experience the world’s smartest AI completely privately on your Apple Mac or Windows computer. You download the app once, and chat forever with no limits on usage, no subscriptions, and 100% privacy. https://t.co/6gWUleN5dd