Proton launches Lumo 2.0 with image generation, advanced reasoning, and improved memory

Proton has released Lumo 2.0, marking the most extensive update to its zero-access encrypted AI assistant since its launch. This version is built on a new architecture that broadens Lumo’s AI capabilities, pushing the product to the frontier of artificial intelligence performance.

On the Artificial Analysis Intelligence Index, Lumo 2.0 demonstrates large gains over its predecessor. The Lite model scores 127% higher and the Max model 240% higher than Lumo 1.4, outpacing previous results in agent work, coding, scientific reasoning, and general knowledge.

The update introduces selectable fast and reasoning modes, enabling users to optimize either for speed in daily tasks or for complex, multi-step problem solving. Everyday queries are now answered up to 76% faster, while challenging prompts activate a visible thinking state that displays Lumo’s reasoning process in real time.

Alongside text, Lumo 2.0 is now multimodal: users can analyze images, generate visuals from prompts or sketches, and edit images within a conversation. Web search capabilities are also improved, with Lumo pulling live results when current information is needed and citing all sources for verification.

Building on this, Lumo 2.0 introduces user-controlled memory and encrypted project workspaces. Users can manage what Lumo learns or forgets, while projects allow secure, multi-session organization of chats, files, and instructions. Custom Lumos further let users design assistants tailored for specific writing or research tasks.