Mistral AI Unveils Large 3, a New Open-Source AI Model Challenging Industry Standards
PARIS – december 3, 2024, 10:28:15 CET - Mistral AI, a Paris-based artificial intelligence company, today announced teh release of Large 3, a significant new open-source AI model designed to compete with established players in the rapidly evolving AI landscape.The launch introduces both instruction-tuned and reasoning versions, offering developers and researchers expanded capabilities for a wide range of applications.
The arrival of Large 3 marks a pivotal moment in the open-source AI movement, providing a powerful and accessible option to proprietary models. This release empowers developers with greater control and customization options, possibly accelerating innovation across industries.The models are intended to lower the barrier to entry for advanced AI development, impacting sectors from customer service and content creation to data analysis and scientific research. Mistral’s commitment to open-source principles positions Large 3 as a key component in the democratization of AI technology.
Mistral reports that the instruction-tuned models achieve performance comparable to similar open-source alternatives while generating significantly fewer tokens, resulting in cost and efficiency gains. The reasoning versions are specifically engineered for complex analytical tasks, offering enhanced capabilities for problem-solving and decision-making. The 14B parameter model demonstrated an 85 percent success rate on the AIME-25 benchmark, a standardized test for evaluating AI reasoning abilities.
Large 3 is currently accessible through Mistral AI Studio and Hugging face. Cloud platform integration includes amazon Bedrock, Azure Foundry, IBM WatsonX, and Together AI, with planned support for Nvidia NIM and AWS SageMaker. Mistral AI confirmed a close collaborative relationship with Nvidia during the model’s development process.
