IT之家 AI

Unsatisfied with Mistral's performance, German search engine Ecosia is betting on Chinese open-source AI models

IT Home, October 8 – According to a report by The Political Newspaper the day before yesterday, German search engine Ecosia recently decided to abandon the French AI model Mistral and switch to a Chinese open-source AI…

Image source · IT之家 AI
Thank you to the netizen from IT Home Huanan Wu Yanzu for sharing the clue!

IT Home On October 8, according to a report by the Political Newspaper the day before, German search engine Ecosia recently decided to abandon the French AI model Mistral and switch to a Chinese open-source AI model.

Christian Kroel, founder and CEO of Ecosia, said in an interview: “We are disappointed with the quality of Mistral. The model is now one whole year behind its competitors.”

In May this year, Ecosia announced that it would abandon its partnership with OpenAI in the United States and switch its AI service provider to Mistral. According to Kroel, this collaboration was not smooth, and the company often encountered technical issues such as server overloads.

According to the original report by IT Home, Ecosia is also concerned about whether Mistral meets environmental goals. In France’s energy structure, nuclear power accounts for a high proportion, and the company questions whether Mistral can remain neutral. Kroll explained: “Mistral relies on funding from international investors, so it is not a true sovereign enterprise.”

This CEO also believes that China is actively promoting the development of open-source large models, which is a great opportunity for Europe. Although Europe missed the first stage of AI model development, now they can use increasingly powerful open-source models without having to invest billions of dollars in self-training of cutting-edge AI.

So now Ecosia has chosen to cooperate with Melious, using the Chinese Qwen, GLM, and Kimi models. This has reduced their costs by half and also improved performance.

Original source

IT之家 AI

Content notes

Original publication and rights belong to the source.

Machine translation · Refer to the original