An ultra-lightweight reasoning model fine-tuned for graduate-level proofs and conversational instruction-following.