French company Mistral has released a new AI model and stated that it outperforms some Chinese counterparts

 • 1944 переглядiв

French startup Mistral has announced the Large 4 model, which will become available on October 27. The company reports strong results in cybersecurity and programming.

On Tuesday, French artificial intelligence startup Mistral unveiled a new AI model, ending a five-month pause between major releases, and said it outperforms many competing open-weight models, reports UNN, citing Reuters.

Details

The three-year-old company, considered the strongest European competitor to U.S. and Chinese AI developers, had not released any new models since May. In August, it began offering customers models from third-party developers, including GLM 5.3 from Beijing-based Z.ai.

"The model we are announcing today outperforms Chinese counterparts on certain metrics — particularly in cybersecurity. So the claim that Europe is unable to compete is not true," CEO Arthur Mensch said during a presentation in Abu Dhabi, without specifying which Chinese models or evaluation criteria he meant.

Mistral said that the Le Chonk model (the official name is Mistral Large 4) will become generally available on October 27.

Before the official release, cybersecurity experts and government agencies will be given access to a version with fewer safety restrictions so they can test the model's capabilities, the company said.

Pierre Stock, Mistral's vice president of science, told Reuters that the model is among the most powerful open-weight systems.

The term "open-weight" means that users can download the model from relevant platforms and run it on their own hardware. Many Chinese models have open weights, while leading U.S. models are usually closed and accessible only through their developers' infrastructure.

Vulnerability was discovered in Meta’s Muse AI agent before launch and hastily fixed - media06.10.26, 12:28

We will add

Stock told Reuters that during testing, the model attempted to break out of its environment, but this was anticipated, and the company managed to prevent such actions.

OpenAI and Anthropic, whose most advanced models demonstrated similar behavior during testing, restricted access to their systems with the highest cybersecurity potential.

After the heads of U.S. AI companies warned last month that artificial intelligence could pose an existential threat to humanity, Mensch dismissed the claims as self-serving.

Mistral, which raised 3 billion euros (3.4 billion dollars) in its latest funding round, presents itself as a safer European alternative. "Closing the gap"

Mistral said that the Large 4 model "closes the gap with frontier models" in areas such as programming, finance, geospatial analysis, manufacturing, and product design.

The company trained this model on its own infrastructure.

AI giant OpenAI carried out a series of dismissals over violations involving "confidential data"02.10.26, 12:22

Stock said that Mistral's goal is to establish "deep" partnerships with its backers, including ASML and new investor Samsung, "across the entire value chain."

He said that Large 4 is capable of designing computer chips, but declined to comment on whether Mistral is currently working with Samsung on chip development.

Popular
News by theme
Protests at French schools have led to the highest number of injuries in decades

 • 2258 переглядiв

Ukraine did not conduct operations in Bulgarian waters — Zelenskyy

 • 2684 переглядiв

Rain will soon return to Ukraine — how many sunny days lie ahead

 • 13910 переглядiв