Both the EU Long Term Programmes for DEMO and the ITER R&D foresee the thermal-mechanical qualification of the beryllium, as neutron multiplier, and lithium orthosilicate or lithium metatitanate as breeder ceramics pebble beds. FZK has performed measurements on the pebble bed thermal-mechanical properties using cylindrical test sections. Using an alternative approach, ENEA, has launched similar testing on the SMARTS mock-up, reproducing on a small scale the reactor reference plane geometry1 instead. The tests have shown that the pebble bed thermal behaviour is strongly affected by the initial filling Packing Factor (PF). In fact, the higher the PF, the higher the thermal conductivity of the bed. Therefore, if the neutron multiplication needs an increase in the pebble PF, the only possibility is to adopt binary pebble beds (small pebbles infiltrating between larger ones) as an alternative to the mono-sized lattice. Using binary pebble beds, the filling quality should be guaranteed against the occurrence of de-mixing or swimming of the larger pebbles over the smaller ones during the thermal transients. A possible solution is to optimise the filling procedure, to improve the PF and its relevant thermal performance, and also to achieve a stable bed lattice during the cycling loads. In this case, the mechanical characteristics of the pebble beds would also be heavily affected, thus requiring a new tests campaign to determine the actual mechanical properties of an “optimised” pebble bed. This paper presents a new filling optimisation method and the experimental results from the compression tests of optimised pebble beds.