What is a known limitation of Claude (and LLMs in general) when performing mathematical calculations?

Large language models including Claude generate mathematical results by predicting tokens, not by performing actual arithmetic operations. This means they can make calculation errors, especially with complex multi-step arithmetic, large numbers, or tricky operations. Claude understands mathematical concepts, notation, and problem-solving approaches well. For accurate calculations, the best practice is to give Claude access to a code execution tool (like Python) and have it write and run code rather than computing mentally. Option A is false — Claude handles math notation and concepts well. Option C invents a specific limitation that doesn't match reality. Option D is wrong — no separate math endpoint exists.