Large Enough
mistral.ai
Today, we are announcing Mistral Large 2, the new generation of our flagship model. Compared to its predecessor, Mistral Large 2 is significantly more capable in code generation, mathematics, and reasoning. It also provides a much stronger multilingual support, and advanced function calling capabilities.

Another day, another model.

Just one day after Meta released their new frontier models, Mistral AI surprised us with a new model, Mistral Large 2.

It’s quite a big one with 123B parameters, so I’m not sure if I would be able to run it at all. However, based on their numbers, it seems to come close to GPT-4o. They claim to be on par with GPT-4o, Claude 3 Opus, and the fresh Llama 3 405B regarding coding related tasks.

benchmarks

It’s multilingual, and from what they said in their blog post, it was trained on a large coding data set as well covering 80+ programming languages. They also claim that it is “trained to acknowledge when it cannot find solutions or does not have sufficient information to provide a confident answer”

On the licensing side, it’s free for research and non-commercial applications, but you have to pay them for commercial use.

  • Couch "Spud" Berner@feddit.nlEnglish
    1·
    2 months ago

    What are the hardware requirements on these larger LLMs? Is it worth quantizing them for lower-end hardware for self hosting? Not sure how doing so would impact their usefulness.