SBS NEWS

France's Mistral Unveils Open-Weight AI Model, Claiming It Outperforms Chinese Competitors


Add SBS News to Google preferred sources
Main image - SBS News

▲ Arthur Mensch, CEO of Mistral AI

French artificial intelligence (AI) startup Mistral unveiled its largest-ever open-weight AI model on October 6, local time.

According to AFP and Reuters, Mistral announced its new open-weight model, "Mistral Large 4," which allows users to download parameters or weights—which determine how calculations are made—and customize them for their own servers.

The product, nicknamed "Chonk," will be released to the general public on October 27.

This marks Mistral's first product release since May.

It is also the company's first announcement since it revealed last month that it raised 3 billion euros (approximately 4.7 trillion won) in a new investment round led by Samsung Electronics, marking the largest fundraising amount ever for a European tech company.

Mistral primarily targets government agencies and corporations that prioritize data security and control over AI through its open-weight models.

In a statement, Mistral said, "From financial services to manufacturing and the public sector, companies and institutions can directly manage their intelligence, data, computing, and operations based on their own core infrastructure without exposing their most valuable knowledge to the outside."

Leading U.S. companies such as OpenAI, Google, and Anthropic use closed model strategies that strictly control how their models are used.

In contrast, Chinese AI companies such as DeepSeek and Moonshot mainly use open strategies.

Arthur Mensch, CEO of Mistral, said at a launch event held in Abu Dhabi, "The model we are announcing today is actually superior to Chinese models in several aspects, including cybersecurity," adding, "The narrative that Europe cannot compete is not true."

(Photo: Getty Images)

※

※ Please note: This article was translated by AI and may contain errors.
Copyright Ⓒ SBS & SBSi. All rights reserved.
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.
Son Gijun View More Articles
AD
AD
AD
AD