CoolFace
Modelpublic

matrixportalx/Llama3-8B-Instruct-Turkish-Finetuned-GGUF

sourceHugging Faceupdated 1y agoView on Hugging Face
3likes454downloads
Model Card

Llama3-8B-Instruct-Turkish-Finetuned GGUF Quantized Models

Technical Details

  • —Quantization Tool: llama.cpp
  • —Version: version: 5162 (2016f07b)

Model Information

Available Files

🚀 Download🔢 Type📝 Description
DownloadQ2 KTiny size, lowest quality (emergency use only)
DownloadQ3 K SVery small, low quality (basic tasks)
DownloadQ3 K MSmall, acceptable quality
DownloadQ3 K LSmall, better than Q3KM (good for low RAM)
DownloadQ4 0Standard 4-bit (fast on ARM)
DownloadQ4 K S4-bit optimized (good space savings)
DownloadQ4 K M4-bit balanced (recommended default)
DownloadQ5 05-bit high quality
DownloadQ5 K S5-bit optimized
DownloadQ5 K M5-bit best (recommended HQ option)
DownloadQ6 K6-bit near-perfect (premium quality)
DownloadQ8 08-bit maximum (overkill for most)
DownloadF16Full precision (maximum accuracy)

💡 Q4 K M provides the best balance for most use cases