Smart home

Mistral Large 4: 1T-parameter model aims to leapfrog rivals

The French AI lab released its largest model yet, promising open weights after safety checks.

Mistral AI released Mistral Large 4 on Tuesday, a new multimodal model with one trillion parameters. TechCrunch reports that the company, backed by ASML and Samsung, aims to offer an alternative to both closed American models and open Chinese ones.

What is Mistral Large 4?

Nicknamed Le Chonk, the model is not yet open-weight. For now, it is accessible only through a public guardrail endpoint. Mistral plans to release the weights in three weeks, after safety testing with trusted partners and governments.

"In the meantime, we'll work with trusted partners and governments to make sure that the open-source weights can be used to defend, but not to [perform] malicious attacks." — Pierre Stock, Mistral VP Science, told TechCrunch.

Security and openness

Enterprise and institutional customers have grown wary of security, Stock said. An open-weight model is easier to audit, he added. Mistral is positioning ML4 as a third way between closed models that can be unplugged and open models often made in China.

Efficiency and performance

Mistral trained ML4 on its own compute using only 4,000 NVIDIA GPUs. Stock told TechCrunch that is two to three times less than Chinese competitors and significantly less than closed-source rivals.

Benchmark results are still pending. Mistral hopes ML4 will be best in class among open-weight models, especially outside China. Thanks to focused training, it could also outperform closed models in specific areas where multimodal capabilities add value.

Where it matters

Stock said optimized use cases include cybersecurity, finance, and chip design. Chip design is core to two of Mistral's main backers: ASML, which led its Series C, and Samsung, which led its Series D last month at a €21 billion valuation (about $24.39 billion).

Bottom line

This is an enterprise model, not something you download on your phone. For consumers, the main question is whether its performance trickles down to smaller models or apps. For now, wait and see what independent tests show and whether the open release happens as promised.

  • AI
  • Le Chonk
  • Mistral
  • Mistral Large 4
  • Nvidia
  • open source

About Lena Fischer

Lena Fischer covers AI features in phones, laptops and apps. She previously spent five years at a Madison software company training and evaluating machine-learning models, including writing the safety guidelines for a customer-service chatbot. She has a BS in computer science from the University of Wisconsin–Madison. She bakes bread most weekends.