What credit score do I need to buy a house, and how do I raise it?

Most people who think their credit is too low to buy a house never call anyone to find out. Just because you think your credit’s bad doesn’t mean your credit is bad. Here is how the credit score for a mortgage actually works: the real minimum by loan program, which of your scores the lender uses, what each step up is worth in dollars, and the fastest honest ways to move your number before you apply.

What credit score do I need to buy a house?

It depends on the loan program, and the program floor is usually lower than people expect. These are the program rules as of November 2026.

Two things sit on top of those rules. First, a lender can set its own floor above the program’s. That’s called an overlay, and many lenders use 620. I work at a mortgage brokerage, so I’m not stuck with one bank’s floor: I can take your file to a lender whose rules fit it. That’s a big part of how a broker differs from a bank.

Second, a lower score usually means more money behind you. On FHA, this is what I see lenders ask for in practice. It’s my experience, not the program rule:

• 640 and up: 3.5% down, no reserves required.
• 580 to 639: 3.5% down, with about two months of reserves.
• Under 580: 10% down and about six months of reserves.

A score in the low 600s is not dead on arrival. Where it hurts most is down payment help, which usually wants a higher score than the loan itself. And sometimes the score isn’t the problem: I’ve seen files where a better score would have fixed nothing because the debt-to-income ratio was the real wall. That math lives on how much house you can afford.

Which credit score do mortgage lenders use?

The middle one. The lender pulls one merged report from all three bureaus (a tri-merge) and takes the middle of your three scores. When two people apply, the loan prices off the lower of the two middle scores.

When one of you is far lower, I run the loan both ways. If the stronger borrower qualifies alone, leaving the other off can save a tier or more of pricing, but that person’s income, their debts and any low down payment option that depended on them leave the file too. They can still help pay the mortgage; they just can’t help qualify for it.

Which scoring model? You don’t pick it, the lender does, and it’s changing. As of November 2026, Fannie Mae and Freddie Mac accept Classic FICO or VantageScore 4.0, and FHA adds VantageScore 4.0 and FICO 10T for case numbers assigned on or after January 1, 2027, with the same minimum scores. All of them still run on a three-bureau report, and the number that counts is the one on the report your lender pulls.

Why is my mortgage credit score different from my app?

Because it’s a different score, often built from data that’s weeks old. Free apps usually show a consumer version of a score, not the mortgage models lenders use, so I never quote off an app. I’ve seen gaps of 30 to 70 points. My view on free score apps: if something is free, you are the product. They earn money when you apply for credit, so treat a nudge like “take this personal loan to boost your score” as a sales pitch, not advice.

Reporting lag causes most gaps. Each creditor reports on its own schedule, usually your statement balance. Pay a card down today and nothing changes until that creditor reports again. You can pay a card in full every month and still report high utilization if the statement closed at a high balance.

Collections weigh more. In my experience collections hit mortgage scores much harder than the app versions. A collection may also show on one bureau and not the others, so your three scores can sit far apart.

You can check the lag yourself. Open your monitoring app, pick a bureau, tap utilization, tap the card and read “last reported.” A card last reported on the 31st refreshes at month end. One last reported on the 6th won’t move until its next cycle. Normally, when it updates on your end, it updates on mine.

A typical case: a borrower saw the mid-700s in an app while the mortgage pull showed 680 to 700. One card had been paid from about $6,000 to about $1,000 and hadn’t reported yet. The fix cost nothing: wait for the statements to cut and check again.

How much does your credit score change your rate and PMI?

More than most people think, and it moves in steps. In my experience every 20 points is a new credit tier, and each tier up improves the rate by about an eighth of a percent at a similar cost. A few points under a line can drop you a whole tier, which is why I check the real middle score before I quote anything. A swing inside a tier, say 736 to 728, doesn’t matter.

The 680 line is the one I watch most on conventional loans, because crossing it changes the price a lot. At the top, pricing stops improving somewhere around 760 to 780, depending on the lender and the loan type.

Mortgage insurance follows the same tiers, and here the prices are published. This is the monthly PMI on a $332,500 loan (a $350,000 home with 5% down) from two national mortgage insurers’ 2026 rate cards, at base rates with standard coverage. A live quote can differ.

Going from 680 to 760 saves about $161 a month, close to $1,930 a year, on PMI alone. Put the rate and the PMI together and a modest jump adds up:

Two exceptions to check before you chase points. Pricing adjustments differ by loan type: on one lender’s rate sheet I analyzed, VA had a single score step, so a VA borrower just above it had nothing to gain from a rescore and a line to fall under. And FHA’s annual mortgage insurance is set by loan size and loan-to-value, not your score, which is one reason FHA can beat conventional for buyers around 680 (FHA vs conventional, side by side). How PMI comes off later is on the mortgage insurance page.

How to raise your credit score fast before you apply

Card balances are the fastest lever, because utilization (your balances divided by your limits) is recalculated every time a card reports. My bands, which are rules of thumb and not lender rules: under 30% total and you’re in good shape; 10% to 20% is the sweet spot while you’re building; past 30% is the danger zone; 50% is a yellow flag; 70% to 75% is a red flag; at 90% and up, card companies start cutting limits.

It works card by card too. The card with the highest utilization moves the score most, so attack it first, and pay before the statement closes, since the statement balance is usually what reports.

Don’t over-clean. A $400 balance on a $20,000 limit doesn’t matter. The same $400 on a $1,000 limit does.

Do not close the account. Not after you pay it off, and not to keep yourself from using it. Closing a card deletes its limit from your total, so the same balances become a bigger percentage. $2,000 spread across $10,000 of limits is 20% used. Close a paid-off card with a $5,000 limit and that same $2,000 is 40%.

Pay revolving debt first. Cards are the lever. In my experience paying off an installment loan, like a car with a few payments left, rarely helps the score and can even dip it, and it spends cash you may need to close. Leave it alone unless its payment is what’s breaking your ratio.

Put every account on autopay for the minimum, scheduled a few days before the due date so a bounced payment can be fixed before it’s late. You can always pay more.

And don’t chase a number you don’t need. If you’re already past the top pricing tier, a personal loan to push you from 760 to 800 buys nothing.

What is a rapid rescore, and is it worth it?

A rapid rescore is the paid shortcut around the reporting lag. After you pay a balance down, the lender sends proof to the bureaus and asks them to update now instead of at the next cycle. It usually takes a few business days to about a week. It fixes balances. It does not fix late payments or collections.

It costs real money. In 2026 I’ve seen about $75 to $130 per account, per bureau, so rescoring every card on all three bureaus adds up fast. I’m selective. Because the lender uses your middle score, updating two of the three bureaus is usually enough, and the third catches up on its own in two to three months. With two borrowers, I rescore the weaker one first and look. If that middle score lands where it needs to, the other person’s accounts may not matter.

It’s worth it when a card is over about 90% used (say $8,000 on an $8,250 limit), when the gain can cross a pricing tier (720 to 740, not 728 to 736), and when you’re under contract with closing a few weeks out. It isn’t worth it when the account is already low, when your app already shows the new balance (then the report has updated), or when you don’t have a contract yet. Two months out, scores update on their own.

Set your expectations: 20 to 40 points is a realistic gain in my files, sometimes 60, and sometimes far less. It’s possible to pay off $10,000 and gain 18 points, especially when old late payments are holding the score down. And watch the re-pull trap: pay a few hundred dollars, gain eight points, pull credit again to check, and the new inquiry takes the points right back.

Should you pay off collections before buying a house?

Not on your own, and not right before you apply. Talk to me first. A paid collection still reports, as paid. It hurts less than an unpaid one, but it’s still a scarlet A, and the cash you’d spend might be exactly what your file needs for closing or reserves. Some collections show on only one bureau, where they may not even touch your middle score.

If paying it is the right move, do it the right way: negotiate, and get the agreement to delete it in writing before you send a dollar. A promise on the phone plus an emailed receipt makes for a weak dispute later.

You also can’t hide the debt. If a collection is on autopay, your bank statements show the payments, and the bureau may report a required minimum higher than what you actually pay. Underwriters read both.

If you have months before you buy, collections and report errors are workable. Disputes take weeks, not days, so start early and keep every letter.

Building credit from a thin file or after a rough patch

A thin file and a damaged file are different problems. A young buyer with one card and a student loan can score in the high 500s with nothing negative on the report. Both get fixed the same way: on-time history and low balances, built on purpose. My ladder for a score in the 600s with a thin or bumpy history:

Done steadily, the low 700s within one to two years is a realistic goal. If your rough patch was a bankruptcy or foreclosure, the waiting periods by program are on buying after bankruptcy or foreclosure. A soft pull lets me show you exactly which accounts your file is missing.

Once you apply, keep your score where it is

Once we submit to underwriting, your credit score is your credit score. From there to closing, the job is to hold it steady: no new accounts, no closed cards, no collection payoffs without a call to me, and autopay on everything (the full list is on what not to do before closing). If you want to know where you really stand, the report I pull is the one that counts, and I’ll tell you exactly what to change, if anything.