xMaulana/FinMatcha-3B-Instruct
xMaulana/FinMatcha-3B-Instruct is a 3.2 billion parameter Indonesian-focused large language model, fine-tuned from the Llama-3.2-3B-Instruct base model. Developed by xMaulana, it excels at understanding and generating Indonesian text, including formal and colloquial speech, while also supporting English for bilingual applications. The model was trained on a wide array of Indonesian datasets, making it particularly adept at handling the nuances of the Indonesian language.
Loading preview...
FinMatcha-3B-Instruct Overview
FinMatcha-3B-Instruct is a 3.2 billion parameter large language model developed by xMaulana, specifically fine-tuned for the Indonesian language. It is built upon the Llama-3.2-3B-Instruct base model and leverages the NekoFi/alpaca-gpt4-indonesia-cleaned dataset to achieve proficiency in Indonesian.
Key Capabilities
- Indonesian Language Proficiency: Optimized for understanding and generating Indonesian text, covering both formal and colloquial styles.
- Bilingual Support: While primarily Indonesian-focused, it also supports English for applications requiring both languages.
- Conversation Handling: Trained to manage various conversational tasks.
- Apache-2.0 License: Available for use under a permissive open-source license.
Performance & Limitations
Evaluations on the Open LLM Leaderboard show an average score of 23.81, with specific metrics including IFEval (0-Shot) at 75.48 and BBH (3-Shot) at 23.19. Users should note that while the model is strong in Indonesian, its performance on non-Indonesian tasks may be limited. Like all LLMs, it may exhibit cultural and contextual biases.
Good For
- Applications requiring robust Indonesian language generation and understanding.
- Bilingual (Indonesian-English) conversational agents.
- Developers looking for a specialized, smaller-scale LLM for Indonesian-centric tasks.