Retro-Engineering a Database Schema: Mistral Models vs. GPT4, LLama2, and Bard (Episode 3)

How do “Mixtral 8x7b” and “Mistral Au Large”, these new LLMs developed by former developers of LLama-2, perform compared to OpenAI, Google, and Meta’s models?

Image generated by the Author with Leonardo.ai

Mistral AI, shaking the LLMs ecosystem

If you have been navigating in the Large Language Models environment lately, you heard about Mistral.ai, a French Start-up founded in April 2023.

After a successful funding round of €385 million in December 2023, it is now valued at nearly €2 billion 🤯 and it has become one of the leading European players in the AI domain.

What explains such success?

First, its three founders are not newbies in the field:

• Arthur Mensch worked at DeepMind and Google’s AI laboratory,
• Guillaume Lample and Timothée Lacroix were employed at Meta.

Needless to say, they leveraged all the experience they had acquired with these major players to quickly develop an open-source alternative to what giants such as OpenAI or Meta were offering.