Latecomer Europe focuses on compact AI models for growth

Arthur Mensch, cofounder and CEO of Mistral AI attends the AI Safety Summit in Bletchley Park, near Milton Keynes, Britain, November 2, 2023. [Photo by Reuters]

Artificial intelligence (AI) startup Mistral AI’s release of its first AI platform is drawing attention as a significant development for European companies, which have lagged in the field of super-sized AI.

According to multiple sources from the information technology (IT) industry, the Paris, France-based company launched a new beta platform service on Monday local time, providing chatbots that generate text based on text commands and embedding.

Based on the Mistral 8x7B AI model, the service is available in three versions, Tiny, Small, and Medium, and aims to offer it at more performance and price points. The current top version, Medium, supports major European languages and codes.

Mistral AI, a latecomer to the industry, aims to penetrate the market with small large language models (sLLM) and open source.

The AI model developed by Mistral AI is classified with only 7 billion parameters, significantly fewer compared to OpenAI’s GPT-3.5 with 175 billion.

Although not universally versatile, sLLM is praised for its ability to perform high-level generation tasks in specific areas.

Mistral AI has recently been in the spotlight in Europe. Founded by former Google DeepMind and Meta researchers in May 2023, the company was valued at $2 billion in the space of just seven months and recently secured an additional investment of 385 million euros.

In France, the company is receiving full support from key figures including President Emmanuel Macron and billionaire Xavier Niel.

It remains to be seen whether Mistral AI will emerge as a new open-source powerhouse, with companies outside the United States that fell behind in the AI competition dominated by companies including OpenAI and Google turning their attention to open source.

German AI startup Aleph Alpha is one such company that is currently developing an open-source model. The company’s AI model, Luminous, is a multi-modal model that can process text in English, German, French, Italian, and Spanish from inputted images.