RN_TR_R2 is now available on Ollama

Advanced Reasoning
for Turkish & English

We build and open-source state-of-the-art language models optimized for instruction-following, multi-step logic, and STEM applications.

Get on Ollama View Models

GRPO Fine-Tuned

Our models leverage Generalized Reinforcement-Preference Optimization for superior multi-step reasoning in complex STEM and cultural tasks.

Bilingual Mastery

Engineered with a Turkish-first approach while maintaining exceptional English comprehension, bridging the language gap in open-source AI.

Optimized Inference

Quantized from 8B parameters down to efficiently run locally on edge devices, macOS, Windows, and Linux via Ollama.

Our Open-Source Models

Discover our flagship reasoning models, heavily fine-tuned on custom datasets.

LATEST

RN_TR_R2

Turkish-language reasoning model upgraded via GRPO on distilled Qwen3 data. Excels at open-ended reasoning, STEM, and cultural questions (82.4% Benchmark).

8B Params GRPO Reasoning
Ollama HF Models

RN_TR_R1

Baseline bilingual LLaMA-based chat model. Optimized for instruction-following, multi-step reasoning, and real-time inference using Unsloth + TRL.

8.03B Params Bilingual Chat
Ollama HF Models

Run Locally in Seconds

bash
# 1. Install Ollama (https://ollama.com/download)
# 2. Run the flagship reasoning model
ollama run RefinedNeuro/RN_TR_R2

# Or use it in your Python applications
from ollama import chat

response = chat(
  model='RefinedNeuro/RN_TR_R2',
  messages=[{'role': 'user', 'content': 'Türkiye\'nin başkenti neresidir ve neden önemlidir?'}],
)
print(response.message.content)