P
PropelGrad

AI Jobs & Internships at Mistral AI 2026

Mistral AI is a French AI company that has rapidly become one of the most respected frontier AI labs in the world through a series of highly capable open-source models — Mistral 7B, Mixtral 8x7B, and subsequent frontier models. Founded by DeepMind and Meta FAIR alumni in 2023, Mistral has raised over €1 billion and is building an alternative to US-dominated AI with a strong European privacy and openness philosophy. Their API and enterprise products compete directly with OpenAI's.

$8,000–$11,000/moIntern monthly pay

AI Roles at Mistral AI

Research Scientist

ML Engineer

Inference Engineer

LLM Fine-Tuning Engineer

Safety Researcher

Platform Engineer

Applied AI Engineer

Enterprise Solutions Engineer

Work Culture at Mistral AI

Mistral maintains the culture of a European research lab — high technical standards, emphasis on publication and open-source contribution, and a work style that is intense but more sustainable than typical Silicon Valley AI startups. The Paris headquarters has a collegiate academic atmosphere with many researchers coming from top French grandes écoles and leading European universities. The company's strong open-source commitment means most research has direct public impact. The relatively small size (300–500 employees) means high individual ownership and direct collaboration with senior researchers.

How to Get a Job at Mistral AI

  • 1.

    Strong French language skills are a plus for Paris-based roles though not required — most technical work is in English

  • 2.

    Contributing to or using Mistral's open-source models and publishing results demonstrates genuine interest

  • 3.

    European work authorization or willingness to relocate to Paris is important for most non-remote positions

  • 4.

    Research roles expect familiarity with Mistral's published work — their mixture-of-experts architecture and efficient training innovations

  • 5.

    The company values efficiency-focused research: demonstrating ability to achieve strong results with limited compute is particularly well-received