unsloth/granite-3.2-2b-instruct
Granite-3.2-2B-Instruct is a 2-billion-parameter instruction-tuned AI model developed by IBM's Granite Team, designed for general instruction-following and thinking capabilities. This model is fine-tuned using a mix of permissively licensed open-source and synthetic data, with a focus on reasoning tasks. It supports a 32768 token context length and is intended for integration into AI assistants across various domains, including business applications, with controllable thinking features. It supports 12 languages including English, German, and Japanese, and excels in tasks like summarization, Q&A, RAG, and code-related functions.
Loading preview...
Model Overview
Granite-3.2-2B-Instruct is a 2-billion-parameter instruction-tuned model developed by the IBM Granite Team, building upon the Granite-3.1 series. Released on February 26th, 2025, this model is specifically fine-tuned to enhance thinking capabilities and general instruction-following. It leverages a diverse training dataset comprising permissively licensed open-source data and internally generated synthetic data tailored for reasoning tasks. A key feature is the controllability of its thinking process, allowing developers to activate it only when necessary.
Key Capabilities
- Enhanced Thinking: Designed to perform reasoning tasks with controllable thought processes.
- General Instruction Following: Capable of handling a wide array of instructions for AI assistant integration.
- Multilingual Support: Supports 12 languages including English, German, Spanish, French, Japanese, and Chinese, with potential for fine-tuning in others.
- Long-Context Processing: Handles tasks requiring extensive context, such as long document summarization and Q&A.
- Diverse Task Performance: Excels in summarization, text classification, extraction, question-answering, RAG, code-related tasks, and function-calling.
Intended Use Cases
This model is well-suited for integration into AI assistants across various domains, particularly business applications where robust instruction-following and reasoning are critical. Its multilingual capabilities also make it suitable for global deployments. The model's ability to toggle its thinking process provides flexibility for different application needs, from direct answers to detailed problem-solving.