Mistral Large 2 Boasts 123 Billion Parameters to Rival Top AI Models
Mistral AI unveils Mistral Large 2, a 123-billion-parameter model that matches GPT-4o and Claude 3 Opus in coding and reasoning tasks while significantly reducing hallucinations.
Mistral AI releases Mistral Large 2, a powerful new AI model featuring 123 billion parameters and a 128k context window. This latest generation pushes the boundaries of cost efficiency and speed, allowing high-throughput inference on a single node. The model supports dozens of human languages and over 80 coding languages, making it a highly versatile tool for developers.
The new model excels in code generation and mathematical reasoning, performing on par with leading industry models like GPT-4o, Claude 3 Opus, and Llama 3 405B. Mistral achieves this by training the model on a massive proportion of code and dedicating significant effort to enhancing its core reasoning capabilities.
A major focus during development is minimizing the model's tendency to hallucinate or generate plausible but factually incorrect information. Mistral Large 2 is fine-tuned to be cautious and discerning, ensuring it acknowledges when it lacks sufficient information to provide a confident answer. The model is available on la Plateforme and released under the Mistral Research License for non-commercial use.