Overview

cnmoro/Mistral-7B-Portuguese is a 7-billion parameter instruction-tuned model, building upon the mistralai/Mistral-7B-Instruct-v0.2 architecture. Its primary goal is to enhance performance and instruction-following capabilities specifically for the Portuguese language. The model was fine-tuned using Unsloth on a dedicated Portuguese instruction dataset.

Key Capabilities & Performance

This model demonstrates strong performance across various Portuguese language tasks, as evaluated on the Open Portuguese LLM Leaderboard. It achieves an average score of 64.7, with notable results in:

• Assin2 RTE: 90.31
• Assin2 STS: 76.55
• HateBR Binary: 79.21
• PT Hate Speech Binary: 68.87

Use Cases

• Portuguese-centric applications: Ideal for chatbots, content generation, and instruction-following tasks requiring high proficiency in Portuguese.
• Sentiment analysis and text classification: Strong scores in Hate Speech and tweetSentBR indicate suitability for these tasks.
• Educational and legal contexts: Performance on OAB Exams and ENEM Challenge suggests potential for applications in these domains, though further evaluation may be beneficial.