The "return to launch site" flight profile flown on December 21, 2015, requires that the first stage re-light three of its engines after separation, turn around, cancel out all its speed, and accelerate back towards the launch site. To do this, it has to keep a lot of fuel in reserve after the separation -- less than you might think, though, because the first stage is quite light once the ascent fuel, second stage, and payload are gone.
In this launch, there were three significant factors that allowed them to do this:
- They "supercooled" the fuel and oxidizer to much lower temperatures than before, which makes them denser, allowing them to carry more fuel in the same volume of tankage on the first stage;
- They lengthened the second stage, allowing more tankage there, so the second stage could provide more of the necessary speed to get the payload to orbit (meaning the stage separation could occur at a shorter distance from the launch site and at lower speed);
- The payload was relatively small, consisting of 11 functional and 1 dummy satellite of about 175kg each -- with mounting and ejection hardware and so on, probably less than 4 tons of payload, compared to the 13 ton payload Falcon 9 is normally capable of launching to LEO.
For full-payload missions, the first stage will have to do more work to push towards orbit, so return-to-launch-site may not be possible, so I believe those missions will use the Florida-to-Atlantic-barge or Texas-to-Florida flight profiles, which require much less fuel in reserve after separation.