ITHome IT Home news, 10/6: On 10/6 local time, French company Mistral AI officially launched the public preview (Public Preview) of the Mistral Large 4 model, abbreviated ML4, codenamed le Chonk.
Starting today, users can experience the preview API on Mistral Studio , and the model weights will be officially opened for download at the end of this month.
Currently, the company is working with top cybersecurity institutions, trusted partners, and government regulators to conduct red-teaming of the model in real business environments. Relevant partners will receive dedicated testing access to the model; the test version appropriately relaxes content safety review policies and has strengthened cybersecurity offense and defense technical capabilities.
ML4 is a natively multimodal model with 1 trillion total parameters and 49 billion activated parameters; Mistral claims its measured performance is not only comparable to the world's top open-source models,but also significantly ahead of all open-weight models developed in Europe and America。
In critical enterprise workloads such as cybersecurity, finance, and law, the model achieves SOTA among all open-source models.
In certain specific areas (such as visual grounding), it goes even further, even surpassing current frontier closed-source models.
According to the company, ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own data center in Europe, and its European deployment is fully operated end-to-end independently by Mistral, strictly complying with EU laws and regulations and not subject to any other third-party digital service provider.
ML4's training data is notably multilingual, with a corpus covering more than 160 natural languages and fully encompassing all official languages of the EU.
In terms of pricing, ML4 costs 1.36 US dollars per million input tokens (ITHome note: approximately 9.1 RMB at current exchange rates) and 4.18 US dollars per million output tokens (approximately 28.1 RMB at current exchange rates).
ITHome attaches the relevant scores below:
Advertising disclaimer: This article contains outbound links (including but not limited to hyperlinks, QR codes, passcodes, and other forms), which are used to convey additional information and save screening time; results are for reference only. All ITHome articles containing external links include this disclaimer.