I totally get where you're coming from! When it comes to LLMs and math, one of the big challenges is how they process numbers as tokens. A number like "1,234,567" might be broken down into smaller parts ("1", ",", "234", etc.), which makes it harder for the model to treat it as a single value. Plus, since LLMs are designed to predict the next token rather than perform precise calculations, they’re not really optimized for mathematical accuracy.

If you need 100% accuracy, I’d really recommend handling the math manually instead of relying on the agent. I know it’s an extra step, but it’ll save you from any potential errors down the line. Let me know if I can help brainstorm a better approach!