It's unlikely that any battery will achieve energy density better than liquid gasoline. The volume of an 18650 cell is pi * (18mm/2)² * 65mm = 16540mm³. The energy density of gasoline is 34.6 J/mm³. So an 18650 filled with gasoline has 572,300 J of energy. One joule is one watt-second, and there are 3600 seconds in an hour, so that's 159 Wh. Current Panasonic NCR18650B cells are 3400 mAh at 3.7V, which is 12.6 Wh. So best case is roughly another factor of 13, barring some sort of nuclear power source.
If we're assuming batteries which use a chemical reaction, then maybe the ultimate limit is a factor of 20 over what we have now (there are chemical reactions more energetic than gasoline burning). If we think in terms of storing charge in ultracapacitors, the limits are probably much higher, depending upon the dielectric constant and the amount of surface area we can squeeze in. Research suggests the
upper limit of a graphene-based supercapacitor is 550 farads per gram. Going by their example a 2 kg capacitor operating at 100 volts, but storing 550 F/gram instead of the 1F/kg of present day supercaps, the capacity would be an incredible 5.5 billion joules of energy, or about 1527 kW-hrs. This equates to 2750 megajoules per kilogram. This is many multiples of the energy from any chemical reaction. Even if a graphene based supercap only achieves 100 F/gram and 30 volts operating voltage, it will still be able to store as much energy per kilogram as gasoline (i.e. the energy in gasoline is about 44 megajoules per kilogram).
On another note however, gasoline is not a great metric to use here because at best gasoline engines turn about 20% of the energy in gasoline into useful work. A more realistic metric then might use ~10 megajoules per kg as the upper bounds for what we can achieve using gasoline as a fuel. Going by that metric, present day batteries are within a factor of three of gasoline, while supercaps are poised to eventually do much better than that.