Mistral Offers Large 4 to Ecosia After Partnership Breakup

Ecosia ends its partnership with Mistral, citing insufficient performance and doubts about sovereignty. Mistral responds by offering immediate access to its new Large 4 model, claiming to narrow the gap with American rivals and announcing a deal with Firefox.

Mistral claims rapid catch-up and partners with Firefox

Mistral has announced a partnership with Mozilla Firefox, allowing one of its models to be offered among the options powering the AI features of the browser. Furthermore, according to Mistral's CEO, the company is now only a few months, rather than years, behind its American competitors. These announcements come as Ecosia has terminated its partnership with Mistral.

Ecosia questions Mistral's sovereignty and mentions nuclear energy

Ecosia questions Mistral's sovereignty by highlighting the presence of non-European investors such as Microsoft, Nvidia, and Samsung. Christian Kroll, CEO of Ecosia, believes that reliance on international funding does not constitute a true act of sovereignty. The company, which claims to have planted 250 million trees, also raises concerns about the significant share of nuclear energy in the French energy mix and its compatibility with its commitments.

Performance: Ecosia criticizes, Mistral highlights Large 4 and customization

Ecosia cites the poor performance of the Mistral models currently in use, namely Mistral small 3.2 and Mistral small 4, with CEO Christian Kroll mentioning a several-year lag behind the competition. According to the ranking by Artificial Analysis, these models are significantly outperformed by Chinese and American competitors. Meanwhile, Mistral recently launched Mistral Large 4, presented as one of the best open AIs and capable of competing with the top Chinese models on certain benchmarks. Guillaume Lample, scientific director and co-founder of Mistral, responded at a press conference by inviting Ecosia to test this new model, even offering immediate early access. He emphasized that creating high-performing models requires time, computing power, and science, and stated that Mistral's growth is exponential. He also highlighted the possibility of improving results by customizing models according to specific needs, a central aspect of Mistral's strategy. Additionally, according to Politico, Ecosia is turning to other open models, including Chinese AIs, while its current AI features still rely on Mistral models.