Mistral Large Instruct 2411 LLM Model

Introduction

Mistral-Large-Instruct-2411 is an advanced large language model with 123 billion parameters, offering state-of-the-art reasoning, knowledge, and coding capabilities. It supports multiple languages and is designed for non-commercial research purposes under the Mistral Research License (MRL).

Architecture

Multi-lingual Support: Handles multiple languages, including English, French, German, Spanish, Italian, and more.
Coding Proficiency: Trained on over 80 coding languages, including Python, Java, and others.
Agent-centric Capabilities: Features enhanced function calling and JSON output.
Large Context Handling: Supports a 128k context window for robust context adherence.
System Prompt: Improved handling for system prompts with a basic instruct template.

Training

Mistral-Large-Instruct-2411 builds on previous versions by improving long context capabilities, function calling, and system prompt handling. The model is optimized for advanced mathematical and reasoning tasks.

Guide: Running Locally

Setup Environment:
- Install vLLM version >= v0.6.4.post1:
```
pip install --upgrade vllm
```
- Install mistral_common version >= 1.5.0:
```
pip install --upgrade mistral_common
```
- Optionally, use a Docker image for setup.
Launch Server:
- Spin up a server with the following command:
```
vllm serve mistralai/Mistral-Large-Instruct-2411 --tokenizer_mode mistral --config_format mistral --load_format mistral --tensor_parallel_size 8
```
- Note: Requires over 300 GB of GPU RAM. Consider using cloud GPUs like AWS or Google Cloud for sufficient resources.
Client Communication:
- Use a Python script to interact with the server, sending and receiving messages to utilize the model's capabilities.

License

Mistral-Large-Instruct-2411 is distributed under the Mistral Research License. Usage is restricted to non-commercial research purposes. For commercial use, a license must be obtained from Mistral AI. The license details can be found here.

More Related APIs