It was. Their models made the assumption that debt ratings were accurate, and they were substituting leveraged mortgage backed securities with AA ratings for risk-free debt. They also underestimated the effects that a liquidity crunch would have - any asset model backed by volatile assets assumes liquidity, but doesn't tell you what will happen in a period of non-liquidity.

Most models assume financial markets are a continuous function, when in reality markets are mostly continuous with periodic step functions. When you hit a period with a step function, the market is either being irrational or correcting prior irrationality. You can't predict these step functions with any accuracy; though many of the big investment banks knew it was coming a year or two in advance. But even if you can predict them, there is no good way to model them.

Goldman Sachs, for all its faults, is run by exceptionally competent people. They came out of the downturn stronger than they went in. They understand the models well because they built most of them. But there were plenty of large market players at Lehman, Merrill, Nationwide, etc. who were ignorant of the problems until it was too late. These are the people who just did what the model told them to because it was making them money, and they didn't understand the risk exposure they really had.