Open4bits / EXAONE-4.0-1.2B-GGUF

This repository provides the EXAONE-4.0-1.2B model converted to GGUF format, published by Open4bits to enable efficient local inference with reduced memory usage and broad CPU compatibility.

The underlying EXAONE model and architecture are developed and owned by LG AI Research. This repository contains only a quantized GGUF conversion of the original model weights.

The model is designed for lightweight text generation and instruction-following tasks, making it suitable for local and resource-constrained environments.

---

Model Overview

EXAONE 4.0 is a large language model family developed by LG AI Research, focusing on strong reasoning, instruction understanding, and multilingual capabilities. This release uses the 1.2B parameter variant, optimized for efficiency while retaining the original architecture.

The GGUF format allows seamless use with popular local inference engines and CPU-based runtimes.

---

Model Details

• Architecture: EXAONE 4.0
• Parameters: 1.2 billion
• Format: GGUF (quantized)
• Task: Text generation, instruction following
• Languages: English, Korean, Spanish
• Weight tying: Preserved
• Compatibility: GGUF-compatible runtimes (CPU-focused inference)

Compared to larger EXAONE variants, this model prioritizes lower memory usage and faster inference, with some trade-off in reasoning depth.

---

Intended Use

This model is intended for:

• Local text generation and chat applications
• CPU-based or low-resource deployments
• Research, experimentation, and prototyping
• Offline or self-hosted AI systems

---

Limitations

• Reduced performance compared to larger EXAONE models
• Output quality depends on prompt design and inference settings
• Not fine-tuned for highly specialized or domain-specific tasks

---

License

This repository follows the original EXAONE license terms as defined by LG AI Research. Users must comply with the licensing conditions of the base EXAONE-4.0-1.2B model.

---

Support

If you find this model useful, consider supporting the project. Your support helps Open4bits continue converting and releasing high-quality, efficient models for the community.