Lumo Review, Pricing & Alternatives (2026)

Lumo is Proton’s hosted AI assistant, launched on July 23, 2025 for web, iOS, and Android. Proton says the Swiss nonprofit Proton Foundation owns it, alongside services used by more than 100 million people. The company says it has not raised venture capital and is building Lumo within Proton’s sustainable business model. Its main audience is people who want general-purpose chat with stricter privacy controls than mainstream assistants. Proton says prompts and replies are not logged after processing, saved Free and Plus histories use zero-access encryption, and chats are not used for model training. Lumo runs models on Proton-controlled servers, so open-source code and models do not make the service self-hostable. Lumo for Business arrived on October 30, 2025, with no minimum team size and a focus on confidential legal, financial, and intellectual-property work.

A typical Lumo session starts in chat, with optional web search, file uploads, and file analysis. Search was off by default at launch, so browsing requires an explicit choice. Projects group conversations and documents for continuing work, while Custom Lumos create assistants around specified tasks. Proton Drive integration brings stored files into that workflow, and image generation is available alongside text chat. Current documentation lists Lumo 2.0 Lite, Lumo 2.0 Max, and Apertus 1.5, with Fast and Thinking modes, while Proton identifies Qwen 3.5, GLM 5.3, Image-Turbo, FireRed-Image-Edit-1.1, and Apertus 1.5 as open-source models running on its servers. The service is delivered through hosted web and mobile applications. Proton describes no public API, BYOK option, or customer-managed inference, leaving Lumo closer to a private assistant than a developer platform or model gateway.

Lumo currently offers Lumo 2.0 Lite, Lumo 2.0 Max, and Apertus 1.5, with Fast and Thinking reasoning modes. Proton says every model is available to users, although paid plans provide substantially higher daily limits for Max. Its privacy documentation identifies Qwen 3.5, GLM 5.3, Image-Turbo, FireRed-Image-Edit-1.1, and Apertus 1.5 as open-source models running on Proton-controlled servers. These implementation models sit behind Lumo’s selectable product labels and should not be assumed to map one-to-one with them. Proton also says Lumo’s code is open source and that the service uses open-source models. The distinction matters: open code and open models do not make Lumo self-hostable. Proton operates the inference infrastructure, and its product, pricing, and model pages describe neither a public API nor customer-managed inference.

The free tier costs $0 and requires no credit card, but it limits Max-model use, messages, chat history, image generations, Projects, web search, and file size. Lumo Plus costs $12.99 monthly, or $9.99 monthly with annual billing at $119.88 per year, and raises those limits while adding unlimited Projects and Custom Lumos. Lumo Professional costs $14.99 per user monthly, or $11.99 per user monthly when billed annually, with a 14-day trial, higher allowances, administrative tools, and compliance features. Workspace Premium with Lumo costs $24.99 per user monthly, or $19.99 monthly annually, includes a 14-day trial, and adds 3 TB per user, custom domains, Mail, Calendar, Drive, Meet, VPN, Pass, and retention policies. Lumo has no public API, BYOK option, or customer-managed inference, so every tier remains a hosted assistant rather than a metered developer service.

Lumo’s strongest distinction is privacy, not breadth. Proton says it deletes prompts and replies after processing, protects saved Free and Plus history with zero-access encryption, and does not use chats to train models, except when users voluntarily submit anonymized feedback on Apertus 1.5. The hosted service runs open-source models on Proton-controlled servers, but Proton does not offer a public API, customer-managed inference, or self-hosting. Independent reviews report slower responses, weaker programming, limited integrations, and less capable creative and image generation than larger assistants. ChatGPT and Claude generally provide broader tools, integrations, and model capability. Perplexity is better suited to citation-first search, while Brave Leo and DuckDuckGo AI Chat keep privacy closer to browser and search workflows. Lumo also lacks the BYOK flexibility of TypingMind, which connects users’ provider keys to multiple model vendors. Choose Lumo for private hosted assistance; choose ChatGPT, Claude, or TypingMind for breadth and control.