MicrosoftAnnouncing the Open Source Big Model Platform Mistral AI A technical cooperation was reached, with a view to strengthening cooperation in the field of artificial intelligence and large language models (LLM).
The three key points of the cooperation include: Microsoft will provide Mistral AI with basic AI computing power through Azure cloud services to help it accelerateModel TrainingAt the same time, Microsoft will provide customers with Mistral AI's basic large models through Azure cloud services, providing users with more choices, including not only OpenAI's GPT series models, but also Mistral AI up to dateThe flagship Mistral Large model released by the company. This cooperation will provide customers with comprehensive services, including deployment and fine-tuning.
Mistral AI, a competitor of OpenAI and a leader in open source large models, raised 490 million euros in 2023 alone.up to dateThe released Mistral Large performs well in mainstream test platforms, second only to GPT-4. It supports 32K contexts, covers multiple languages such as English, French, Spanish, and German, and has excellent performance in mathematics, reasoning, and coding. The model is currently available on Microsoft's Azure cloud service.
In addition to large-scale model collaboration, Mistral AI is also based onup to dateMistral AI was founded in 2023. The three co-founders have extensive experience in the field of large models and have worked at Facebook's parent company Meta Platforms and Google's AI research lab DeepMind. Their project LLaMA is an influential open source project.maximumIt is one of the projects of the Linux Foundation, and has spawned multiple open source projects, such as Alpaca, Guanaco, LuoTuo, Vicuna, Koala, etc.
Mistral AI's earlier Mistral7B isStrongestOne of the open source large language models, its performance is better than Llama213B in all benchmarks, and even surpasses or is comparable to Llama134B in most tests. This technical cooperation with Microsoft will further consolidate Mistral AI's leading position in the field of large models, providing users with more choices and high-quality services.