Mistral AI releases Mistral Large 2 as an answer to Meta and OpenAI's latest models

Mistral AI has unveiled Mistral Large 2, the latest iteration of their flagship large language model. This new model offers significant advancements over its predecessor in areas such as code generation, mathematics, and reasoning. It also boasts enhanced multilingual support and advanced function calling capabilities.

Key features of Mistral Large 2 include a 128k context window and support for dozens of languages like French, German, Spanish, Italian, Portuguese, Arabic, Hindi, Russian, Chinese, Japanese, and Korean. Additionally, it supports over 80 coding languages, including Python, Java, C, C++, JavaScript, and Bash. Designed for single-node inference with long-context applications, Mistral Large 2's 123 billion parameters allow it to operate at high throughput on a single node.

Performance-wise, Mistral Large 2 not only surpasses its predecessor but also competes with leading models such as GPT-4o, Claude 3 Opus, and Llama 3 405B. A significant portion of the development effort was focused on improving the model's reasoning capabilities and reducing the likelihood of generating hallucinations — plausible-sounding but incorrect or irrelevant information. The model has been fine-tuned to be more cautious and discerning, ensuring more reliable and accurate outputs. Additionally, it is trained to acknowledge when it lacks sufficient information to provide a confident answer.