A language model trained specifically to excel at one task or domain (like mathematics) rather than performing well across many different tasks.
Multi-step reasoning, logic puzzles, mathematical problem-solving