Mistral comes out swinging for enterprise AI customers with new Le Chat Enterprise, Medium 3 model

Mistral has started the French AI’s start raising a load of private financing, but it has not yet broken the first AI’s use plans in the world, especially when it comes to adopting the institution and developers.

But this may change starting today: The company unveiled the Le Chat Enterrise, a unified auxiliary platform of artificial intelligence designed for productivity and privacy on the scale of institutions, supported by its new model 3 Mediterranean 3, which surpasses larger tools in a small part of the cost (here, “Akbar” indicates the number of parameters, or internal model settings, which usually indicate more capabilities, but also.

https://www.youtube.com/watch?

Le chat entity is a ChatgPT assist

Available on the web and through mobile phone applications, Le Chat Enterrise is similar Rough A group of different applications and sources of data.

It is designed to unify the functions of artificial intelligence in one environment and a privacy that allows deep allocation, functional function, and rapid publication.

Among the main features that will be important for business owners and technical decision makers ::

• Google Drive, SharePoint, Gmail, and more, without exposing or launching information to third parties)
• Document libraries with automatic capabilities and martyrdom
• Certified conductors and agents to automate tasks without a symbol
• Integration of specialized models and memory -based allocation
• Hybrid publishing options with the support of the public cloud, private VPCS, and local hosting

Le Chat Enterprise supports smooth integration in the tools and current workflow. Companies can build artificial intelligence agents designed for their processes and maintain full sovereignty on publishing and data-without locking the seller.

The platform privacy structure is committed to strict access controls and supports the registration of full audit, ensuring data governance for organized industries. Institutions also profit fully control of the AI ​​Strong intelligence-from infrastructure and the features of the statute to customization at the level of the model and user’s interfaces.

Given the general suspicion of some Western companies and governments across China and its growing library of strong models open source from companies there, along with the Mistral site in the European Union and the narrow data protection laws, it must be followed (the public data protection regulation also known as GDP and EU AI law) and the new Mistral Foundation. Companies).

Mistral also offers improvements to LE Chat Pro plans and team plans, targeting individuals and small teams looking for productive tools supported by their language models. All levels benefit from the basic capabilities provided in Le Chat Enterprise.

Mistral Medium 3 surpasses GPT-4O and even Claude 3.7 Sonnet on the main standards and is available via API and on-Prem

Mistral Medium 3 introduces a new performance layer in the company model assortment, which is placed between lightweight and wide -ranging models. It is a ownership model, and this means that unlike previous versions of a mediator, it is not available under an open source license and must be used through Mistral and faced the application programming interface or its partners.

The model, designed for institutions, provides more than 90 % of the standard performance of CLAUDE 3.7 Sonnet, but at the cost-0.40 dollars per million input symbols and $ 20.80 per million output symbols, compared to $ 3 of Sonatat/15 dollars per input/output.

The criteria indicate that Mistral Medium 3 is particularly strong in software development tasks. In coding tests such as Humaneval and MultipL-E, it corresponds to or exceeds both Claude 3.7 Sonnet and Openai’s GPT-4O.

According to human assessments by a third party, Llama 4 MAVERICK excels in 82 % of coding scenarios, and the matter is more than 70 % of cases.

The model also works competitive through languages ​​and methods. Compared to Llama 4 MAVERICK, a higher victory rates in English (67 %), French (71 %), Spanish (73 %), Arabic (65 %), decrease in multimedia performance with noticeable degrees in tasks such as Docvqa (0.953), AI2D (0.937), and Cartqa (0.826).

Mistral Medium 3 has been improved for the integration of the institution. It supports mixed and installed publishing, provides post -training, and easily connects to business systems. According to Mistral, it is already used in the experimental version by organizations in sectors such as financial services, energy and health care in the workflow of the energy field and solutions facing customers.

Mistral Medium 3 can now be reached via Mistral’s La Platforme Api, Amazon Sagemaker, with support soon to IBM Watsonx, Nvidia Nim, Azure Ai Foundry and Google Cloud Vertex.

Meanwhile, Le Chat Enterprise is available in the Google Cloud market, and will soon be launched on Azure Ai and Aws Bedrock.

For those ready to explore the assistant experience, Le Chat is available on Chat.mistral.ai, as well as in the application store and the Google Play store, with no credit card required to start.

By combining a highly efficient model with a customizable institution platform, Mistral AI pushes a concerter to reduce barriers that prevent the adoption of developing and privacy artificial intelligence in the world of institutions.