tools
EXAONE
EXAONE is LG AI Research’s multimodal model family for text, image, document, reasoning, and industry-focused AI applications.
EXAONE 4.5 combines language processing with a proprietary visual encoder, allowing it to process text and images together. It is designed for complex document understanding, reasoning, Korean language and visual contexts, and professional industry tasks.
Researchers and developers can access EXAONE 4.5 as an open-weight model through Hugging Face. Commercial use, hosted access, and API availability may depend on licensing or LG’s partners; no official public pricing page was found.
Features
- Processes text and images in a native multimodal architecture
- Uses an integrated proprietary visual encoder
- Understands complex documents with tables, formulas, and multi-column layouts
- Uses multi-token prediction to improve inference speed
- Supports Korean language and Korean visual-context understanding
- Provides reasoning for mathematical, scientific, and programming tasks
- Supports agent and tool-use evaluation scenarios
- Available as an open-weight model through Hugging Face
Use cases
- Analyze technical and business documents
- Extract information from charts, tables, and visual documents
- Answer questions about images and Korean visual contexts
- Generate or review code and solve technical problems
- Build industry-specific document and reasoning applications
- Evaluate multimodal models for research and development