The Paris-based startup Mistral AI has introduced a cutting-edge large language model (LLM) named Mistral Large, claiming superior performance in multitasking language understanding assessments. The company’s announcement highlighted its prowess in mathematical and coding evaluations, positioning it as a potential rival to existing industry giants. Notably, GPT-4 was not included in the comparative study.
Advancements in AI Competency
Mistral Large, which has not been compared with other recent models such as xAI’s Grok or Google’s Ultra, represents an advancement from the company’s earlier offerings. Guillaume Lample, the company’s founder, has expressed confidence that the new model surpasses previous iterations. Additionally, Mistral AI unveiled Le Chat, an AI chat interface that operates on top of their models, akin to ChatGPT’s relationship with GPT-3.5 and GPT-4.
December reports revealed that Mistral Large, backed by major investors like Nvidia and Salesforce, boasts fluency in over 20,000 English words and multiple European languages. Its linguistic versatility demonstrates the company’s commitment to creating a multilingual model.
Mistral AI’s Market Position and Partnerships
While the original Mistral AI model was open-source, Mistral Large has shifted to a proprietary model, similar to those by OpenAI. This change has prompted mixed reactions from the tech community. The previous version, Mistral Medium, secured a sixth-place ranking on Chatbot Arena, a site that evaluates AI chatbots through a Bradley-Terry model for predicting competitive outcomes.
Mistral AI has also gained industry buzz with the announcement of a partnership with Microsoft, integrating Mistral Large into Azure AI Studio and Azure Machine Learning platforms, potentially expanding its accessibility and application.