French founder of artificial intelligence startup Mistral AI, Arthur Mensch, attends the Viva Technology show at Parc des Expositions Porte de Versailles in Paris, France, on May 22, 2024.
Chesnot | Getty Images Entertainment | Getty Images
French artificial intelligence firm Mistral is on Tuesday launching its first reasoning model to compete with rival options from the likes of OpenAI and China’s DeepSeek.
“We’re announcing in a couple of hours our new reasoning model, which is very much competitive with all the others and has the specificity of being able to reason in multiple languages,” CEO Arthur Mensch told CNBC’s Arjun Kharpal onstage during a fireside chat at London Tech Week.
Reasoning models are systems that can execute more complicated tasks through a step-by-step logical thought process. Mistral’s new model “is great at mathematics [and] great at coding,” according to Mensch.
Other reasoning models on the market currently include OpenAI’s o1, which was released to users of its ChatGPT service late last year and Chinese AI lab DeepSeek’s R1.
Mistral, which is backed by U.S. tech giant Microsoft, specializes in so-called open-weight large language models. These are systems that make their underlying “parameters” — elements that are adjusted during training to improve performance — publicly available.
This essentially enables developers to access and modify the model’s core knowledge and bypass the high costs and time associated with training a system from scratch.
Mistral’s CEO said that the unique selling point of its upcoming Magistral reasoning model is that it will be able to reason with European languages. “Historically, we’ve seen U.S. models reason in English and Chinese models reason in Chinese,” he said.
While it’s focusing on European languages for now, Mensch added that Mistral will add support for more languages later down the line.
At the start of this year, Chinese AI startup DeepSeek released a reasoning model called R1 that shocked the AI community — and global markets — promising competitive performance with OpenAI’s rival o1 model at a lower cost.
