"Thermoelectric efficiency depends on the figure of merit, ZT. There is no theoretical upper limit to ZT, and as ZT approaches infinity, the thermoelectric efficiency approaches the Carnot limit. However, no known thermoelectrics have a ZT>3."[1]
There seem to be some in promising materials with a ZT of 2.2[2] which is around 20% similar to a solar cell. It seems again if more R&D where applied we might be able to make that cheap and practical.
1. https://en.wikipedia.org/wiki/Thermoelectric_materials#Devic...
2. https://newatlas.com/most-efficient-thermoelectric-material/...
2. Quantum tunneling - electrons will tunnel back to lower charge density region, and preventing them from doing so is effectively impractical with modern day tech.
If they were any actual thermocouples with 20% efficiency, they would've nuked piston engines long time ago.
The maximum what you can get can't be higher than the hypothetical ideal quantum diode made from a material pair.
There are much more down to earth alternatives that beat both stirlings and thermocouples on reliability and specific power density: thermionic converters, thermoaccoustic generators, AMTEC converters, radiophotovoltaic (works only in 0g)