Ecosia Replaces France's Mistral with Chinese Open-Source AI Models
German search engine Ecosia has abandoned France's Mistral AI to adopt Chinese open-source AI models including Qwen, GLM, and Kimi through a partnership with Melious. This transition reportedly slashed Ecosia's AI operational costs by 50% while improving service performance and stability. This shift highlights a growing trend of European tech companies adopting competitive Chinese open-source large language models over regional offerings. It demonstrates how high-performing open-weight models allow businesses to avoid multi-billion-dollar training costs while achieving superior cost-efficiency. Ecosia CEO Christian Kroll criticized Mistral for lagging a full year behind competitors and causing frequent server overloads after Ecosia initially switched to it from OpenAI in May. Additionally, Ecosia raised concerns regarding Mistral's environmental impact due to France's nuclear energy reliance and questioned its sovereignty status given its international investor base.
## BACKGROUND
Qwen is a family of open-weight large language models developed by Alibaba Cloud, while GLM (General Language Model) was developed by Tsinghua University and Z.ai. Kimi is another prominent Chinese model series developed by Moonshot AI for context-heavy and multimodal tasks. These Chinese AI initiatives actively push open-source capabilities, providing European firms with powerful alternatives to proprietary Western models.