CoolFace
Modelpublic

UgurI/Qwen3.8-27B-Turkish-CPT-GGUF

sourceHugging Facecc-by-nc-sa-4.0updated 1mo agoView on Hugging Face
0likes64downloads
Model Card

Qwen3.8-27B-Turkish-CPT-GGUF (Q4KM)

GGUF quantizations for Qwen3.8-27B-Turkish-CPT, adapted for high-quality Turkish comprehension, reasoning, and instruction following.

Model Description

  • —Base Model: Qwen/Qwen3.8-27B (Text-only causal LM backbone)
  • —Quantization: Q4_K_M (15.41 GB, 4.92 BPW)
  • —Context Length: Up to 262,144 tokens
  • —Compatibility: Ollama, LM Studio, llama.cpp, Jan, OpenWebUI

Available Quantizations

FileQuant MethodFile SizeDescription
Qwen3.8-27B-Turkish-CPT-Q4_K_M.ggufQ4_K_M15.41 GBRecommended balance of speed, memory (~18 GB VRAM/RAM), and Turkish accuracy

Benchmark & Exam Performance

MetricBase Model (Qwen3.8-27B)Turkish-CPT-GGUFImprovement
Turkish Validation Perplexity7.58386.9360-8.55%
Turkish Test Perplexity8.05347.3303-8.98%
TurkishMMLU (90-Q Stratified)60.00%75.56%+15.56%
Belebele Reading Comprehension (TR)83.33%90.00%+6.67%
Belebele Reading Comprehension (EN)96.67%96.67%100% Retained

Quick Start with Ollama

  1. 1.Download the Qwen3.8-27B-Turkish-CPT-Q4_K_M.gguf and Modelfile files.
  2. 2.In your terminal, run:
bash
   ollama create qwen3.8-turkish -f Modelfile
   ollama run qwen3.8-turkish

Quick Start with LM Studio / llama.cpp

  1. 1.Place Qwen3.8-27B-Turkish-CPT-Q4_K_M.gguf into your LM Studio models directory.
  2. 2.Or run directly with llama-cli:
bash
   llama-cli -m Qwen3.8-27B-Turkish-CPT-Q4_K_M.gguf -p "Türkiye'de yenilenebilir enerji kaynaklarının önemi nedir?" -n 256

Citation & Attribution

  • —Training Corpus: VNGRS-AI (vngrs-ai/vngrs-web-corpus)
  • —Base Model: Qwen Team (Qwen/Qwen3.8-27B)
  • —Author: Ugur Inanc