PhysOrg News has this piece...
Turning heat to electricity... efficiently

...which indicates 40% conversion efficiency (of the Carnot Limit) of heat to electricity, but with a theoretical ceiling of +90% of the Carnot Limit. Potentially that's really good news for space systems, especially RTGs and ultimately "solid state" nuclear reactors.

The real problem with all thermal systems, in space, is getting a decent temperature differential to begin with - when heat has to be dumped via radiative cooling, then efficiency is inherently limited. A solid state thermoelectric system can't avoid that issue, but it can get rid of a whole lot of power conversion machinery inefficiencies and failure modes. Thus why it's used even with its currently pathetic 5-8% of Carnot Limit efficiency. Pumped up to 40%, then ion/plasma-drive probes become viable, like the ESA's Pluto Orbiter from a few years ago.