Mistral Large rivals GPT-4 and Claude 2: Setup, Testing, and Function Calling

Mistral Large is the world’s second-ranked model generally available through an API according to official announcement; have a look at the performance comparison on MMLU (Measuring massive multitask language understanding):

However, if you include both Gemini Ultra 1.0 (83.7%) and Gemini Pro 1.5 (81.9%), they slightly outperform Mistral Large.

Let’s have a look at some of Mistral Large’s characteristics:

• Multilingual support for English, French, Spanish, German, and Italian.
• Extended 32K context window for more granular response generation.
• Supports function calling and JSON mode to streamline interactions with internal code and databases, facilitating complex applications.
• 81.2% on MMLU, closely behind GPT-4, Gemini Ultra and Gemini Pro
• Not open, only available on Mistral or Azure

In this article, I will walk you through

• Setting up a local environment with Mistral Large
• Mistral AI Model Pricing
• Running initial tests with Mistral Large
• Function calling with Mistral Large
• Further resources to dig deeper