Gemma-4-E2B-NoAudio-GGUF
This repository provides GGUF weights for the text-only version of Gemma-4-E2B-it. The audio components have been removed for maximum efficiency on home hardware.
Key Features
Imatrix Optimized: All files are quantized using a custom Importance Matrix (imatrix) generated from 1,000+ chunks of Thai text to ensure superior reasoning and language quality.
Home Hardware Ready: Highly efficient and lightweight, perfect for running locally on consumer GPUs and CPUs.
Available Files
- Q8_0 High fidelity for maximum accuracy.
- Q4_K_M Best balance of speed and intelligence.
Usage
This model is compatible with most GGUF runners. For the best experience at home, we recommend:
- LM Studio
- Ollama
- llama.cpp / llama-server