Mistral’s Large 4 open weights arrive with a 1-point coding lead over GLM-5.3: “There’s nothing you can’t do with this model that you can do with competitors.”
On Thursday, Mistral released the weights for its frontier open model, Mistral Large 4 (ML4), on Hugging Face. Made available The post Mistral’s Large 4 open weights arrive with a 1-point coding lead over GLM-5.3: “There’s nothing you can’t do with this model that you can do with competitors.” appeared first on The New Stack .
On Thursday, French AI company Mistral unveiled the open weights for its frontier model, Mistral Large 4 (ML4), on Hugging Face. The model has been under development for three weeks and is touted as the "best open agentic coding model in the world," surpassing GLM-5.3 on coding benchmarks, though slightly trailing in cybersecurity tests.
Mistral's VP of Science, Pierre Stock, explained that the company aims to be in a position to dictate the frontier, emphasizing the importance of open weights for transparency, auditability, and business continuity. Mistral's VP of Science stressed that living in a world without open weights is unacceptable, drawing parallels to the potential consequences of power outages.
With the release of ML4, Mistral aims to demonstrate the capabilities of an open-weight model that excels in coding and cybersecurity, while being more cost-effective and efficient than competitors.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
