Proton pushes the Lumo 2.0 assistant to compete directly against ChatGPT and Claude

The Reddit user community has detected the change in the AI ​​model by noticing the update of the Lumo privacy support page. It details the large open source language models (LLM) that power Lumo, and that the current versions used by Lumo 2.0 are Qwen 3.5 – Alibaba’s LLM -, GLM 5.2, Image-Turbo and FireRed-Image-Edit-1.1.

A week ago, Proton announced Lumo 2.0, which brings with it advanced models, thinking mode, multimodal capability, image analysis, a new look, new memory features, and better web search. However, the biggest benefit is in its performance, since if Lumo 2.0 Lite has become 127 percent more powerful, Lumo 2.0 Max does so by 240 percent in the Artificial Analysis Intelligence Index.

In fact, according to internal tests from the company itself, its founder and CEO, Andy Yen, maintains that “users can no longer perceive a qualitative difference between Lumo 2.0 Max and the latest models from OpenAI and Anthropic,” according to the specialized media Cybernews.

However, in the announcement, Proton did not mention at any time the move to Qwen 3.5 and the abandonment of Mistral, the pride of European AI according to the European think tank (CEPS), a model that Proton has practically used since the launch of Lumo 1.1 in August 2025, after the debut of the AI ​​assistant in July 2025.

The leap in performance and capabilities entails a geopolitical turn, since Qwen, having been developed by Alibaba in China, comes pre-trained with alignment filters that condition certain responses with political biases. According to a recent study by the CEIAS center, Qwen is one of the Chinese models with the least political bias detected, far below Kimi or DeepSeek.

In terms of performance, the Qwen models have outperformed the Mistral in virtually all open weight benchmarks. This, for the Lumo user, means true multimodality, advanced reasoning, and a larger context window (with the Memory feature).

Another aspect is that the data does not travel to China, since Proton has downloaded the Qwen open weights and installed them on its own European server infrastructure. That is, the processing continues to be protected by Proton policy and the data is not stored on third-party platforms.

With this new move, Lumo continues to be Proton’s AI assistant that is characterized by its zero-access encryption, its no-logs policy (it does not save or share chats with third parties), the absence of training with user data (unlike other chatbots with AI) and a 100% European infrastructure. But now take advantage of the power offered by the Qwen models.