Model Overview

MaziyarPanahi/calme-3.1-llamaloi-3b is a 3.2 billion parameter language model, building upon the meta-llama/Llama-3.2-3B architecture. Its primary differentiation lies in its specialized fine-tuning for the French Legal domain, making it particularly adept at processing and generating content relevant to French law.

Key Capabilities

• French Legal Domain Specialization: Enhanced performance for tasks within the French legal context.
• Llama-3.2-3B Base: Leverages the foundational strengths of the Llama-3.2-3B model.
• Extended Context Window: Supports a context length of 32768 tokens, allowing for processing longer legal documents or complex queries.
• Quantized GGUF Availability: Offers quantized GGUF models for efficient deployment and inference.

Performance & Usage

While a small model, it has been evaluated on the Open LLM Leaderboard, achieving an average score of 24.01. It uses the ChatML prompt template for interaction. Users should be aware of its small size, which may lead to sensitivity to hyperparameters and potential performance variations for some prompts. Ethical considerations regarding potential biases and limitations, common to large language models, are also noted, recommending safeguards and human oversight in production environments.