This model is a Gemma 3 4B variant distilled from xAI’s Grok 3. It was fine-tuned to emulate Grok’s depth and structured clarity, particularly in tasks involving complex thought, such as problem-solving, coding, and mathematics.
This dataset consists of 100 high-quality Grok 3 completions responding to deep questions, solving math problems, and writing or analyzing code. The aim was to distill Grok’s analytical approach and technical versatility into a smaller, accessible model.
This Gemma model was trained 2x faster with
Unsloth and Huggingface's TRL library.
