How different are pricing models such as the binomial model for us options from the systems used by large financial companies? For example, http://finance.bi.no/~bernt/gcc_prog/ has wonderful material, but would a bank or a hedge fund ever think of using the code at that site?Thanks.

I'm not sure what you're asking. The code at the site seems useful and good quality, no doubt many quants have borrowed it for their own routines. Certainly a lot of quant programmers have "Numerical Recipes in C" on their desks.No large financial company would knowingly get their fundamental source code from an Internet site, however individual programmers might cut and paste.The routines themselves seem to be designed only for general hypothetical investigation. For example, the bond pricing routine assumed a constant interest rate and periodic payments. Pricing a real bond requires computing the precise cash flows on the precise days, and constructing the appropriate discount factor for each flow. Moreover, it would need to handle features like amortization, call, cross-currency features and so forth.

Textbook models often fail to produce or reproduce market prices because they typically fail to capture important institutional details like settlement lags in the underlyings, or even more basic things like day counts and such. One has to take these models and adapt them to the relevant market. Currency options are a good example. The textbook Black Scholes model will be close, but not quite there. Short-term bond options using the Black model are another - the market convention for computing the forward bond price is probably slightly different than a professor would come up with.

thanx, very helpful. I wasn't sure how different "class room material" is from what actual quants do.