Mistral-Large-2402

Mistral Large is an advanced language model developed by Mistral AI that excels at text generation and complex multilingual reasoning tasks. It was launched on February 26, 2024 and is available through Mistral's own platform "la Plateforme" as well as through Microsoft Azure.

Key features:
- Generates coherent, contextually relevant text in multiple languages including English, French, Spanish, German and Italian
- Performs well on reasoning and knowledge benchmarks like MMLU, HellaSwag and WinoGrande
- Shows top performance on coding and math tasks evaluated through HumanEval and GSM8K benchmarks
- Supports a 32K token context window for precise information recall from large documents
- Enables function calling to interface with user tools, APIs or databases
- Provides a JSON format output mode to facilitate structured data extraction

How it works:
- Accessible to developers through Mistral's "la Plateforme" hosted in Europe for building applications and services
- Integrated with Azure AI Studio and Azure Machine Learning for a seamless user experience
- Can be self-deployed in user environments for sensitive use cases, with access to model weights
- Enables application development requiring advanced text generation and reasoning
- Supports large-scale tech stack modernization through function calling and constrained output
- Allows design of effective moderation policies through precise instruction-following

Integrations:
- Microsoft Azure: Azure AI Studio and Azure Machine Learning
- Mistral's "la Plateforme" for hosting and access
- Self-deployment for integration with internal systems and environments

Use of AI:
Mistral Large uses generative AI to perform a wide range of tasks spanning text generation, complex reasoning and code generation.

AI foundation model:
Built on advanced AI foundation models enabling high performance on various benchmarks.

How to access:
- Available as a web app, API, and for self-deployment
- Suitable for developers, businesses, and technology enthusiasts seeking advanced AI capabilities
- Offers open-weight endpoints with competitive pricing, but the model itself is not open source