Microsoft-Stunder AI Lab Mistral Debuts Reasoning model to compete openly

The French artificial intelligence company Mistral is launching its first reasoning model on Tuesday to compete with rival options from OpenAi and China's Deepseek.

“We will announce our new reasoning model within a few hours, which is very competitive with everyone else and the specificity to be able to reason in several languages,” CEO Arthur Mensch told Arjun Kharpal on the cnbc stage during a heat chat during a Hekhat during the London Tech Week.

Reasoning models are systems that can perform more complicated tasks via a step -by -step logical logical thinking process. The new Mistral model “is great in mathematics [and] Great in coding, “according to Mensch.

Other reasoning models on the market are currently OpenAI's O1, which was released at the end of last year to users of the Chatgpt service and the R1 of Chinese AI Lab Deepseek.

Mistral, which is supported by the American tech giant Microsoftspecializes 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 get access to and the core knowledge of the model and to bypass the high costs and time that are accompanied by training a system completely again.

The CEO of Mistral said that the unique sales argument of the upcoming magisterial reasoning model is that it will be able to reason with European languages. “Historically we have seen our models in English and Chinese models in Chinese,” he said.

Although it focuses on European languages ​​for the time being, Mensch added that Mistral will add support for more languages ​​later.

At the beginning of this year, the Chinese AI startup Deepseek released a reasoning model with the name R1 that the AI ​​community -and global markets -shocked, promised competition performance with OpenAI's rival O1 model at lower costs.