| Property | Value |
|---|---|
| Parameter Count | 7.94B |
| License | Apache 2.0 |
| Base Model | Qwen2-7B-Instruct |
| Paper | View Paper |
| Supported Languages | 39 languages |
What is Pangea-7B?
Pangea-7B is a groundbreaking fully open-source Multilingual Multimodal Multicultural Large Language Model (LLM) that represents a significant advancement in global AI accessibility. Built on the LLaVA-NeXT architecture with a Qwen2-7B-Instruct backbone, this model was trained in 2024 using the comprehensive 6M PangeaIns dataset.
Implementation Details
The model leverages advanced tensor processing with BF16 precision and implements sophisticated multimodal capabilities through the LLaVA-NeXT framework. It can process both text and images, making it versatile for various applications.
- Built on Qwen2-7B-Instruct architecture
- Supports both text-only and multimodal inputs
- Implements efficient context handling up to 2048 tokens
- Uses specialized tokenization for multiple languages
Core Capabilities
- Multilingual support for 39 diverse languages including Asian, European, and African languages
- Multimodal processing combining text and image inputs
- Advanced prompt handling with system message support
- Flexible generation parameters for temperature and sampling
Frequently Asked Questions
Q: What makes this model unique?
Pangea-7B stands out for its comprehensive multilingual support across 39 languages and its ability to process both text and images in a unified framework, making it one of the most versatile open-source models available.
Q: What are the recommended use cases?
The model excels in multilingual applications, image-text understanding tasks, cross-cultural communication, and general language processing across diverse linguistic contexts. It's particularly useful for applications requiring global reach and multimodal capabilities.