How much does elevation gain slow you down in a marathon?

Enough that where the climbing sits matters more than how much there is. Grade-adjusted pacing gives a per-kilometre answer for your specific course.

There is a widely repeated rule of thumb that every 10 metres of climb costs a handful of seconds, and it is not useless — but it hides the variable that matters most, which is where the climbing falls. Three hundred metres of gain spread evenly across a course is a very different race from 300 metres concentrated between 30 and 35 km.

The reason is compounding. The metabolic cost of a climb is roughly the same wherever it appears, but your ability to pay that cost is not: on depleted glycogen with accumulated fatigue, the same gradient produces a much larger pace loss and a much longer recovery afterwards. That is why courses are better characterised by their hardest late kilometre than by their total ascent.

The honest way to answer the question for your own race is to run the actual elevation profile through a gradient-cost model at your goal time. That produces a per-kilometre target that spends more time on the climbs and less on the descents, and it tells you which specific splits are going to be the slow ones before you are standing in them.

One caveat about the data. Elevation figures derived from digital elevation models carry noise of roughly 5–10 metres on flat courses, so a genuinely flat urban marathon can read 20–30 feet higher than it is. Treat small differences in total ascent between flat courses as measurement noise rather than as a meaningful ranking.

Glossary: GAP (Grade-Adjusted Pace), Minetti Gradient Cost