I'm afraid if you want a quality answer, you're going to have to dig deeper into the numbers than you or any of the commenters have yet. Get comfortable and take your time. It's worth it.

Last time I bought, about two years ago, I worked out the marginal cost of credit for all of my loan options, 3 each from 3 different lenders. There are different ways to do this, but for me, the most straightforward approach was to start with the smallest down payment, calculating the annual cost of credit for that loan, then work out the annual cost savings for each incrementally higher down payment, and turn that into an annual return percentage based on the incremental money down.

Our basic options were 3.5% down (FHA w/ PMI), 10% down w/ PMI, and 20% down, no PMI. I'll describe my calculations only for the lender I actually used, the one with the lowest closing costs, but I did this for each of 3. I don't have all the exact numbers with me but I remember enough to reconstruct it reasonably well. Our home cost $122,900, FTR.

With the 3.5% down FHA loan, the rate was higher and the PMI was nearly $200/mo - and it didn't go away no matter how much you paid the loan down. So, unsurprisingly, it was worth paying more to reach the next tier. But I did the math to make sure (and you should do the math!). The extra $8k to reach 10% down eliminated roughly $1800 in annual costs, for a return of over 20%. However, below 90% LTV, the rate didn't change much, and the PMI was only $32/mo. So, putting another $13K into the down payment would have only saved $768 per year: 6.25% pretax, 4.7% after tax (25% bracket). On top of that, we'll easily be out of PMI very soon, dropping the long-term marginal cost of credit even lower.

Meanwhile, our market returns have run in the 15% range - not without risk, but totally acceptable risk during an accumulation phase. The investment returns from year 1 alone covered all of the PMI we'll ever pay. Does this mean 10% is the best DP for you? Absolutely not! It just illustrates that the details matter.