Kimi K3, developed by Moonshot AI, is currently one of the highest-ranked models on the Artificial Analysis Intelligence Index (score: 57), placing it among the top 5 LLMs worldwide. Integrating it into Lumo 2.0 Max would significantly strengthen the range of available models.

Arguments in favour of this integration:

• Superior performance — Its Intelligence Index score of 57 surpasses the majority of direct competitors and usefully complements GLM-5.2 (score: 51). This model diversity would enable finer routing based on the nature of each request.

• Extended context — With a native context window of 1 million tokens, Kimi K3 offers capabilities equivalent to GLM-5.2 for processing long documents, which is essential for academic research and the analysis of complex reports.

• Alignment with Proton's strategy — Moonshot AI has expressed its intention to publish the model's weights (2.8T parameter version), which aligns perfectly with Proton's commitment to open-source and transparency.

• Complementarity with existing models — Kimi K3 excels particularly in agentic tasks and long-horizon reasoning, areas where its combination with GLM-5.2 would enable intelligent routing tailored to each use case.

Recommendation: Integrate Kimi K3 into the selection of models available through the Lumo 2.0 Max interface, while maintaining the optimised automatic routing or allowing manual selection by the user.

Expected impact: This integration would improve Lumo's competitiveness against competing solutions while respecting Proton's confidentiality and Swiss respectively European infrastructure principles. Professional users would benefit from enhanced response quality for demanding tasks without compromising data protection.