I am interested to read about GPSDOs. Rarely do people state what they want to use it for, how accurate it needs to be, will it be on continuously or turned on when needed, where is the antenna likely to be located. I have been experimenting with various very cheap configurations and the best figures are like - worst deviation of OCXO 10MHz to GPS 1pps is ±50ns or half a cycle, worst deviation from 10MHz is ±1mHz. And because it is locked to the GPS signal, will long term be accurate. Just changing the OCXO these figures can be improved. Using a dual frequency GPS will also improve it. More attention to circuitry will improve it. I have been reading about Johnny B Good and his attempts to eliminate diurnal fluctuations from his Rubidium standard, with the conclusion the deviations are most likely due to the GPS signal, not the Rb. I don't believe JBG has a reason to do this other than pursuing perfection. He may have to buy a dual frequency GPS, the only justification being he can.
Some people do have a genuine need. Long baseline interferometry needs accurate time. Radio transmissions in the 10s of GHz need accurate frequency. The figures I gave above would probably satisfy someone broadcasting at 10GHz, with an error of 1Hz. But the LBI requirement could be less than 5ns, so 50ns won't be adequate.
So without a requirement or a budget, answering 'What GPSDO do you recommend' is nigh impossible.