Taklaxbr/gemma3-4b-turkish-thinking

VISIONConcurrent Unit Cost:1Model Size:4.3BQuant:BF16Context Size:32kPublished:Jun 13, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

The Taklaxbr/gemma3-4b-turkish-thinking model, originally developed by HarunYigit and archived by VeriPazarı, is a 4.3 billion parameter Gemma3-4B language model specifically fine-tuned for Turkish reasoning and instruction following. It excels at generating Chain-of-Thought style responses in Turkish and is optimized for inference via Ollama and llama.cpp in GGUF format. This model is designed for applications requiring logical reasoning and stable final answer formats in Turkish.

Loading preview...

Overview

Taklaxbr/gemma3-4b-turkish-thinking is a 4.3 billion parameter Gemma3-4B model, originally developed by HarunYigit and archived by VeriPazarı, that has been fine-tuned to enhance its Turkish reasoning capabilities. It focuses on generating Chain-of-Thought (CoT) style responses and following instructions in Turkish. The model is released in GGUF format, making it optimized for efficient inference using Ollama and llama.cpp.

Key Capabilities

  • Turkish Reasoning: Specialized in generating logical, step-by-step (Chain-of-Thought) responses in Turkish.
  • Instruction Following: Fine-tuned to accurately follow Turkish instructions.
  • Stable Output Format: Designed to produce consistent and evaluable "final answer" formats.
  • Hardware Friendly: Optimized for local inference on personal devices using Ollama and llama.cpp.
  • Benchmark Suitability: Structured to be suitable for multiple-choice benchmark tests.

Good For

  • Applications requiring advanced logical reasoning in Turkish.
  • Developing chatbots or AI assistants that need to provide detailed, thought-out responses in Turkish.
  • Local deployment on consumer hardware for Turkish NLP tasks.
  • Research and development in Turkish AI, particularly for evaluating reasoning abilities.

Finetunes
1 models